INFRASTRUCTURE PRICING / USD30 products in this category/Research snapshot Nov 3, 2026

02 / AGENT RUNTIMES

Agent runtimes

Which parts of the bill are separate?

Platform + execution + models

The three parts of the bill

Inspect runtime provider

01 / PLATFORM

Not published / month starting price

No Managed Agents recurring free allowance stated; separate eligible Claude subscription API credits may apply.

02 / EXECUTION

Usage charges

$0.08 per running session-hour plus model tokens; the Managed Agents API remains beta.

Usage rates

Running session$0.08/session-hour, metered to the millisecond
Web search$10/1,000 searches when used

03 / EXTRAS

Models & other costs

Model input/output tokens and prompt caching are charged separately. US-only inference applies a 1.1x token multiplier; fast mode can add a premium.

The $0.08 session rate excludes model tokens. Idle, rescheduling and terminated time do not accrue runtime; this is not the Agent SDK by itself.

Plans & sources

COMPARISON

Compare products

30 of 30 products

Search productsPricing statusShortlist 0

Sort by

Published runtimes pricing. Entry fees and compute estimates are not total bills. Unknown values sort last.
Shortlist	Product	Platform entry	Execution & published rates	Models & other extras	Operational tradeoff
	Claude Managed AgentspreviewSourceNov 3, 2026Runs Claude agent sessions with a managed harness and execution environment.	Not publishedMonthly starting price, not total	$0.08 per running session-hour plus model tokens; the Managed Agents API remains beta. Running session$0.08/session-hour, metered to the millisecond Web search$10/1,000 searches when used	Model input/output tokens and prompt caching are charged separately. US-only inference applies a 1.1x token multiplier; fast mode can add a premium.	The $0.08 session rate excludes model tokens. Idle, rescheduling and terminated time do not accrue runtime; this is not the Agent SDK by itself.
Plans and sources for Anthropic / Claude Managed Agents PRICE RECORD / claude-managed-agents Anthropic / Claude Managed Agents Runs Claude agent sessions with a managed harness and execution environment. Status: previewChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Managed Agents beta $0.08/running session-hour plus model token rates Managed agent harness and code execution environment Beta header required; token, web-search and residency modifiers can add cost Usage meters Running session$0.08/session-hour, metered to the millisecond Web search$10/1,000 searches when used Free allowance / trial No Managed Agents recurring free allowance stated; separate eligible Claude subscription API credits may apply. Additional costs Model input/output tokens and prompt caching are charged separately. US-only inference applies a 1.1x token multiplier; fast mode can add a premium. Watch for The $0.08 session rate excludes model tokens. Idle, rescheduling and terminated time do not accrue runtime; this is not the Agent SDK by itself. Resource and session limits Isolation Managed agent sandbox and harness Persistence Session log survives idle periods; execution runtime bills only while running Regions Data residency options depend on platform; US-only inference has a token multiplier GPU Not published Maximum session Not published Concurrency (running at once) Not published Workload fit Claude-specific autonomous agents that need hosted sessions. Source record Managed Agents pricingNov 3, 2026 Hosted product overviewNov 3, 2026
	Amazon Bedrock AgentCore RuntimepublicSourceNov 3, 2026Runs agent code in managed microVM sessions or managed EC2 instances.	$0Monthly starting price, not total	microVM v2: $0.1276/vCPU-hour consumed and $0.0169/GB-hour memory; v1: $0.0895 and $0.00945. Instances add a management fee to EC2. v2 consumed CPU$0.1276/vCPU-hour v2 memory$0.0169/GB-hour	Model inference, S3 or ECR artifacts, network transfer, CloudWatch and optional AgentCore services are separate.	CPU charges follow used CPU time. Memory charges follow peak use each second. The two microVM versions have different rates.
Plans and sources for AWS / Amazon Bedrock AgentCore Runtime PRICE RECORD / aws-agentcore-runtime AWS / Amazon Bedrock AgentCore Runtime Runs agent code in managed microVM sessions or managed EC2 instances. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. microVM v1 $0.0895/vCPU-hour plus $0.00945/GB-hour Per-second consumption billing; 1-second minimum 128 MB minimum billed memory; up to 8-hour sessions microVM v2 $0.1276/vCPU-hour plus $0.0169/GB-hour; committed baseline rates $0.0997 and $0.0132 where available Consumption billing with idle memory reclamation 1-second minimum; 128 MB minimum billed memory Instances EC2 On-Demand cost plus 12% management fee; 7.8% for supported G-series GPUs Persistent EC2-backed sessions in the customer account Up to 14-day sessions; EC2, EBS and transfer bill separately Usage meters v2 consumed CPU$0.1276/vCPU-hour v2 memory$0.0169/GB-hour Free allowance / trial No recurring Runtime allowance stated; new AWS customers may receive general Free Tier credits. Additional costs Model inference, S3 or ECR artifacts, network transfer, CloudWatch and optional AgentCore services are separate. Watch for CPU charges follow used CPU time. Memory charges follow peak use each second. The two microVM versions have different rates. Resource and session limits Isolation microVM session or managed EC2 instance Persistence Instances support persistent sessions; microVM sessions end at termination Regions See AWS regional availability GPU Instances support selected GPU types; microVM does not Maximum session 8 hours microVM; 14 days Instances Concurrency (running at once) Quota-dependent Workload fit Agent code that needs isolated sessions and AWS integrations. Source record Official pricingNov 3, 2026 Compute typesNov 3, 2026
	Workers and AgentspublicSourceNov 3, 2026Runs agent logic in Workers and stateful agents in Durable Objects.	$0Monthly starting price, not total	Workers Free has 100,000 requests/day; Workers Paid starts at $5/month with 10 million requests and 30 million CPU-ms included. Workers requests$0.30/million over included Paid requests Workers CPU$0.02/million CPU-ms over included Paid CPU Durable Object requests and duration$0.15/million requests and $12.50/million GB-seconds beyond allowance	Durable Object duration and SQL storage, Workers AI model usage, KV, R2 and other bindings can add charges.	Agents is not a separate free hosting plan. Worker, Durable Object and model meters can all apply; Pages is outside this record.
Plans and sources for Cloudflare / Workers and Agents PRICE RECORD / cloudflare-agents Cloudflare / Workers and Agents Runs agent logic in Workers and stateful agents in Durable Objects. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Workers Free $0/month 100,000 requests/day; 10 ms CPU per invocation Durable Objects have separate free limits Workers Paid $5/month plus overage 10 million requests and 30 million CPU-ms/month Up to 5 minutes CPU per invocation; Durable Objects separate Usage meters Workers requests$0.30/million over included Paid requests Workers CPU$0.02/million CPU-ms over included Paid CPU Durable Object requests and duration$0.15/million requests and $12.50/million GB-seconds beyond allowance Free allowance / trial Workers Free has daily requests and CPU limits; Durable Objects also publish separate free limits. Additional costs Durable Object duration and SQL storage, Workers AI model usage, KV, R2 and other bindings can add charges. Watch for Agents is not a separate free hosting plan. Worker, Durable Object and model meters can all apply; Pages is outside this record. Resource and session limits Isolation Workers isolate; stateful Agent maps to a Durable Object Persistence Durable Object storage is separate from Worker execution Regions Cloudflare global network; data-location controls vary by feature GPU Not published Maximum session Worker CPU limit applies; long waits use stateful primitives Concurrency (running at once) No fixed agent concurrency published Workload fit Lightweight stateful agents near users on Cloudflare. Source record Workers pricingNov 3, 2026 Agents limitsNov 3, 2026 Cloudflare plansNov 3, 2026
	CrewAI CloudpublicSourceNov 3, 2026Deploys CrewAI agentic workflows to CrewAI-hosted infrastructure.	$0Monthly starting price, not total	Basic free with 50 workflow executions/month; Enterprise custom with sized executions and overage. See plan details for usage rates.	Model and connected-tool costs are not specified as included on the pricing page.	Only Basic and custom Enterprise tiers are published; do not invent a mid-tier execution rate.
Plans and sources for CrewAI / CrewAI Cloud PRICE RECORD / crewai-cloud CrewAI / CrewAI Cloud Deploys CrewAI agentic workflows to CrewAI-hosted infrastructure. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Basic $0 2 automations and 50 workflow executions/month on CrewAI Cloud Maximum 50 executions; no published overage Enterprise Custom Cloud, customer VPC or customer infrastructure Custom executions and flexible overage by quote Usage meters Not separately published Free allowance / trial 50 workflow executions recur monthly on Basic. Additional costs Model and connected-tool costs are not specified as included on the pricing page. Watch for Only Basic and custom Enterprise tiers are published; do not invent a mid-tier execution rate. Resource and session limits Isolation CrewAI Cloud or customer deployment Persistence Deployment history and tracing listed Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) Not published Workload fit CrewAI workflows needing a hosted visual and API runtime. Source record PricingNov 3, 2026
	Agent Bricks Agent RuntimepublicSourceNov 3, 2026Deploys custom agent servers on managed Databricks Apps compute.	Not publishedMonthly starting price, not total	Agent Runtime uses Databricks Apps DBUs: Medium 0.5, Large 1, XLarge 3 DBU/hour while running; USD per DBU depends on cloud, region and plan. App compute0.5, 1 or 3 DBU/hour by size while running	Model Serving, foundation-model inference, database, storage and network services can add charges. The Databricks DBU USD rate varies by cloud and region.	Agent Bricks is a platform label. This record prices the current Databricks Apps Agent Runtime, not legacy Model Serving or Databricks general hosting.
Plans and sources for Databricks / Agent Bricks Agent Runtime PRICE RECORD / databricks-agent-runtime Databricks / Agent Bricks Agent Runtime Deploys custom agent servers on managed Databricks Apps compute. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Free Edition evaluation $0 for qualified personal use within fair-use quota Up to 3 Databricks Apps No reliability SLA; not a production allowance Medium App 0.5 DBU/hour while running; regional USD rate varies Up to 2 vCPUs and 6 GB RAM Provisioned capacity billed while app runs Large App 1 DBU/hour while running; regional USD rate varies Up to 4 vCPUs and 12 GB RAM Provisioned capacity billed while app runs XLarge App 3 DBU/hour while running; regional USD rate varies Up to 12 vCPUs and 48 GB RAM Provisioned capacity billed while app runs Usage meters App compute0.5, 1 or 3 DBU/hour by size while running Free allowance / trial Free Edition is a quota-limited personal evaluation product, not a production monthly credit. Additional costs Model Serving, foundation-model inference, database, storage and network services can add charges. The Databricks DBU USD rate varies by cloud and region. Watch for Agent Bricks is a platform label. This record prices the current Databricks Apps Agent Runtime, not legacy Model Serving or Databricks general hosting. Resource and session limits Isolation Managed Databricks App runtime Persistence App configuration persists; running compute bills while active Regions Databricks Apps regional availability varies by cloud GPU Not listed for Databricks Apps Agent Runtime Maximum session Not published Concurrency (running at once) Horizontal app scaling available; numerical cap depends on workspace Workload fit Custom agent servers tied to Databricks data and governance. Source record Agent deployment stackNov 3, 2026 App compute sizesNov 3, 2026 Free Edition limitsNov 3, 2026
	DBOS CloudCustom quoteSourceNov 3, 2026Hosts durable workflow applications with autoscaling and managed recovery.	Not publishedMonthly starting price, not total	Contact sales; official docs say CPU time is billed, but publish no rate. See plan details for usage rates.	Model tokens and external services are separate; exact Cloud billing terms are unpublished.	Do not call the open-source library's $0 price the hosted Cloud price.
Plans and sources for DBOS / DBOS Cloud PRICE RECORD / dbos-cloud DBOS / DBOS Cloud Hosts durable workflow applications with autoscaling and managed recovery. Status: quoteChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. DBOS Cloud Quote; CPU-time unit rate unpublished Application hosting, autoscaling and workflow recovery Contact sales required Usage meters Not separately published Free allowance / trial No hosted Cloud allowance verified; open-source DBOS is a separate self-hosted product. Additional costs Model tokens and external services are separate; exact Cloud billing terms are unpublished. Watch for Do not call the open-source library's $0 price the hosted Cloud price. Resource and session limits Isolation Managed cloud executor Persistence Durable workflow history and recovery Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) Not published Workload fit DBOS applications needing managed recovery. Source record Cloud deployment docsNov 3, 2026
	Dify CloudpublicSourceNov 3, 2026Hosts Dify-built agent, chatflow and workflow apps with web and API access.	$0Monthly starting price, not total	Sandbox free; Professional $59/month or $590/year; Team $159/month or $1,590/year per workspace. See plan details for usage rates.	Message credits are model-dependent. After they run out, use a model-provider key; provider token charges are separate.	Message credits are not a fixed number of model calls across models. Dify open source is separate from Cloud.
Plans and sources for Dify / Dify Cloud PRICE RECORD / dify-cloud Dify / Dify Cloud Hosts Dify-built agent, chatflow and workflow apps with web and API access. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Sandbox $0 200 message credits, 5 apps, 50 MB knowledge storage 1 member; 10 knowledge requests/minute Professional monthly $59/workspace/month 5,000 message credits/month, 50 apps, 5 GB knowledge storage 3 members Professional annual $590/workspace/year Same tier; annual billing Do not treat as $59 monthly charge Team monthly $159/workspace/month 10,000 message credits/month, 200 apps, 20 GB knowledge storage 50 members Team annual $1,590/workspace/year Same tier; annual billing Annual commitment Enterprise Custom Private deployment options Quote Usage meters Not separately published Free allowance / trial Sandbox's 200 message credits are published without a clear monthly reset; do not assume recurring credits. Additional costs Message credits are model-dependent. After they run out, use a model-provider key; provider token charges are separate. Watch for Message credits are not a fixed number of model calls across models. Dify open source is separate from Cloud. Resource and session limits Isolation Managed Dify Cloud app Persistence Knowledge storage and log history vary by tier Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) Knowledge request rate published; agent concurrency not published Workload fit No-code or low-code hosted agent apps. Source record Cloud pricingNov 3, 2026 Annual pricingNov 3, 2026
	Harness RuntimepublicSourceNov 3, 2026A managed agent session adds tools and pause/resume to an isolated sandbox.	$0Monthly starting price, not total	$0.044 per consumed vCPU-hour and $0.0095 per allocated GB-hour at standard rates; Agent Droplets plans can discount eligible usage. CPU$0.044/consumed vCPU-hour Memory$0.0095/allocated GB-hour Full-use shape examples1/1 $0.0535, 2/2 $0.107, 2/4 $0.126, 4/8 $0.252, 16/32 $1.008 per hour	Session volumes, snapshots, checkpoints and custom templates: $0.05/GiB-month each. Public egress: $0.01/GiB. Action Gateway calls and inference are separate.	CPU is consumed time, not provisioned time; memory, storage, tools, and model calls remain separate.
Plans and sources for DigitalOcean / Harness Runtime PRICE RECORD / digitalocean-harness-runtime DigitalOcean / Harness Runtime A managed agent session adds tools and pause/resume to an isolated sandbox. Status: publicChecked: Nov 3, 2026 Selected estimate basis, before extras Plan / regionStandard usage Medium 2 vCPU / 4 GB / Not specified CPU billingActive CPU time Required monthly plan$0 Minimum spend / included usage$0 / $0 $0.044 / CPU-hour + $0.0095 / GB-hour Metered compute $5.56 Required plan + $0 Applied usage credit - $0 Minimum adjustment + $0 Estimated month $5.56 Published Medium fixed Harness shape. CPU bills on actual consumption; 4 GB memory bills on allocation. No Agent Droplets discount. Storage, egress, tools and inference excluded. Compute + storage + outbound (before other costs) All three costs included$6.26 / month Excludes tax, model APIs, and charges outside this scope. Budget plan: Standard usage Medium 2 vCPU / 4 GB / Not specified. Check unverified session or concurrency limits CPU estimate uses your utilization; memory stays allocated Only standard configurations, with no recurring credit. Managed Agents balances fund compute, storage and egress. Agent Droplets discounts and promotions are excluded. Monthly storage is prorated over a 744-hour model month. Only compute, retained storage and outbound traffic are modeled. Models, tax and other services are excluded. Budget line items Required monthly plan $0 Standard usage Medium 2 vCPU / 4 GB Compute usage before credit $5.56 Published Medium fixed Harness shape. CPU bills on actual consumption; 4 GB memory bills on allocation. No Agent Droplets discount. Storage, egress, tools and inference excluded. Retained storage $0.50 Source You pay for the disk space held by session volumes. Funding examples are not free allowances. The 744-hour month is for comparison, not the provider's billing formula. Count extra copies separately. Outbound traffic $0.20 Source Public-internet outbound traffic bills per GiB; no allowance is deducted. Recurring credit applied $0 Shared across verified compute, storage and outbound charges; applied once. Minimum-spend adjustment $0 Applied after recurring credit. Published plans Other plans and regional prices may be available at the source. Standard usage Usage-based; no subscription required XSmall through XLarge sessions A usable account balance is needed for charges Agent Droplets Pro $50/month 15% off eligible Managed Agents usage; fee funds $50 subscription balance, covering $58.82 of standard-rate eligible usage Third-party models are not discounted; plan is in public preview Agent Droplets Team $200/month 20% off eligible usage; fee funds $200 subscription balance, covering $250 of standard-rate eligible usage Third-party models are not discounted; plan is in public preview Usage meters CPU$0.044/consumed vCPU-hour Memory$0.0095/allocated GB-hour Full-use shape examples1/1 $0.0535, 2/2 $0.107, 2/4 $0.126, 4/8 $0.252, 16/32 $1.008 per hour Free allowance / trial No recurring free compute grant confirmed; a $5 new-customer promotion is shown as one-time. Agent Droplets balances are funded by their plan fees. Additional costs Session volumes, snapshots, checkpoints and custom templates: $0.05/GiB-month each. Public egress: $0.01/GiB. Action Gateway calls and inference are separate. Watch for CPU is consumed time, not provisioned time; memory, storage, tools, and model calls remain separate. Resource and session limits Isolation Hardware-isolated microVM session Persistence Pause/resume/fork; retained storage keeps billing Regions Not published on pricing page GPU Not published Maximum session Not published Concurrency (running at once) Not published Workload fit Agent sessions that use managed tools and pause/resume. Source record Official Harness Runtime pricingNov 3, 2026
	Gemini Enterprise Agent Platform Agent Runtime (formerly Vertex AI Agent Engine)publicSourceNov 3, 2026Deploys containerized agent code on Google's managed agent runtime.	$0Monthly starting price, not total	$0.085 per allocated vCPU-hour and $0.009 per GiB-hour RAM; 1-year and 3-year flexible savings rates differ. Agent Compute$0.085/allocated vCPU-hour on demand Agent Memory$0.009/allocated GiB-hour on demand Agent Storage$0.30/GiB-month	Model tokens, Memory Bank, Sessions, Agent Gateway, Skill Registry and related storage or operations can bill separately.	The old $0.0864 Agent Engine rate is superseded by the current $0.085 Agent Compute rate. GiB and GB are not interchangeable.
Plans and sources for Google Cloud / Gemini Enterprise Agent Platform Agent Runtime (formerly Vertex AI Agent Engine) PRICE RECORD / google-agent-runtime Google Cloud / Gemini Enterprise Agent Platform Agent Runtime (formerly Vertex AI Agent Engine) Deploys containerized agent code on Google's managed agent runtime. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. On demand $0.085/vCPU-hour plus $0.009/GiB-hour RAM Per-second resource billing; idle time between prompts is not billed Model, memory bank, sessions, gateway and storage are separate 1-year flexible savings $0.0765/vCPU-hour plus $0.0081/GiB-hour RAM Published committed-use rate Requires eligible savings-plan commitment 3-year flexible savings $0.068/vCPU-hour plus $0.0072/GiB-hour RAM Published committed-use rate Requires eligible savings-plan commitment Usage meters Agent Compute$0.085/allocated vCPU-hour on demand Agent Memory$0.009/allocated GiB-hour on demand Agent Storage$0.30/GiB-month Free allowance / trial Pricing page includes 50 Agent Compute vCPU-hours and 100 Agent Memory GiB-hours per account each month before paid rates; model tokens are separate. Additional costs Model tokens, Memory Bank, Sessions, Agent Gateway, Skill Registry and related storage or operations can bill separately. Watch for The old $0.0864 Agent Engine rate is superseded by the current $0.085 Agent Compute rate. GiB and GB are not interchangeable. Resource and session limits Isolation Managed container Persistence Sessions and Memory Bank are separately metered services Regions See Google Cloud Agent Runtime locations GPU Not published for Agent Runtime Maximum session Long-running operations up to 7 days Concurrency (running at once) Configurable instances; quotas apply Workload fit Containerized agents that use Google Cloud models and managed state. Source record Current platform pricingNov 3, 2026 Agent Runtime overviewNov 3, 2026 Rename release noteNov 3, 2026
	Hatchet CloudpublicSourceNov 3, 2026Hosts task orchestration while application workers connect from customer compute.	$0Monthly starting price, not total	1 million runs and 1 million events/month free; then $30/million runs and $5/million events on paid usage. Runs$30/million after monthly allowance Events$5/million after monthly allowance Logs/traces storage$0.0004/GB-hour after 10 GB free	Customer-hosted worker compute and model tokens are separate; users beyond 50 are $10/user.	A run may spawn other billable runs. Failed runs count; retries do not add a run.
Plans and sources for Hatchet / Hatchet Cloud PRICE RECORD / hatchet-cloud Hatchet / Hatchet Cloud Hosts task orchestration while application workers connect from customer compute. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. No-card free $0 2,000 runs/day Hard cap; tasks drop after free limit Pay as you go $0 base; metered overage 1 million runs and 1 million events/month; 100 workers; 50 users 7-day retention; 10 GB/day active storage Custom Quote Dedicated capacity, multi-region and enterprise controls 100 million+ runs/month positioning Usage meters Runs$30/million after monthly allowance Events$5/million after monthly allowance Logs/traces storage$0.0004/GB-hour after 10 GB free Free allowance / trial 1 million runs and 1 million events recur monthly with card-on-file PAYG; no-card tier has 2,000/day limit. Additional costs Customer-hosted worker compute and model tokens are separate; users beyond 50 are $10/user. Watch for A run may spawn other billable runs. Failed runs count; retries do not add a run. Resource and session limits Isolation Managed orchestrator; workers run separately Persistence 7-day retention on PAYG Regions AWS Oregon, N. Virginia and Ireland; Sydney private beta GPU Worker-dependent Maximum session Not published Concurrency (running at once) 100 workers on PAYG; task concurrency not published Workload fit High-volume durable task dispatch with self-hosted workers. Source record PricingNov 3, 2026
	Inngest CloudpublicSourceNov 3, 2026Orchestrates durable agent steps and can run code through separately priced Inngest sandboxes.	$0Monthly starting price, not total	Free 50,000 executions/month; Pro from $99/month for 1 million; Business from $499/month for 10 million. Optional sandbox CPU$0.0396/vCPU-hour Optional sandbox memory$0.0108/GB-hour	Hosted sandbox compute is separate; customer-hosted function workers and model tokens are not covered by orchestration price.	Execution counts include each step and the function call. Hosted sandboxes cost extra.
Plans and sources for Inngest / Inngest Cloud PRICE RECORD / inngest-cloud Inngest / Inngest Cloud Orchestrates durable agent steps and can run code through separately priced Inngest sandboxes. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Free $0/month 50,000 executions, 500,000 events, 5 concurrent steps 3 workers; 24-hour trace history Pro From $99/month 1 million executions, 5 million events, 100+ concurrent steps 20 workers; 7-day trace history Business From $499/month 10 million executions, 50 million events, 500 concurrent steps 100 workers; 14-day trace history Enterprise Custom Custom executions and limits Quote Usage meters Optional sandbox CPU$0.0396/vCPU-hour Optional sandbox memory$0.0108/GB-hour Free allowance / trial 50,000 executions and 500,000 ingested events recur monthly on Free. Additional costs Hosted sandbox compute is separate; customer-hosted function workers and model tokens are not covered by orchestration price. Watch for Execution counts include each step and the function call. Hosted sandboxes cost extra. Resource and session limits Isolation Orchestration plus optional sandbox isolation Persistence Durable steps; trace retention 24 hours to 90 days by tier Regions Not published on pricing page GPU Not published Maximum session Sandbox maximum 24 hours on Pro/Business Concurrency (running at once) 5 free, 100+ Pro, 500 Business concurrent steps Workload fit Event-driven agents that need retries and step state. Source record PricingNov 3, 2026
	LangSmith DeploymentpublicSourceNov 3, 2026Hosts LangGraph agent servers as serverless or dedicated deployments.	$39Monthly starting price, not total	Plus $39/seat/month; one small serverless deployment included; further runtime and database resources are metered. Runtime compute0.0675 LSU/vCPU-hour; 1 LSU = $1 Runtime memory0.0090 LSU/GiB-hour Database compute and memory0.177 LSU/vCPU-hour plus 0.025 LSU/GiB-hour	Tracing and model gateway usage can add charges; production deployments have database uptime costs.	Do not merge this with LangSmith Sandboxes. The free Developer seat cannot deploy agents.
Plans and sources for LangChain / LangSmith Deployment PRICE RECORD / langsmith-deployment LangChain / LangSmith Deployment Hosts LangGraph agent servers as serverless or dedicated deployments. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Developer $0/seat/month Tracing; no Deployment access One seat; not a hosted runtime plan Plus $39/seat/month plus resource usage One small serverless deployment and 10,000 base traces/month Additional serverless or dedicated deployments charged on resource time Enterprise Custom; annual invoice Cloud, hybrid and self-hosted options Contract pricing Usage meters Runtime compute0.0675 LSU/vCPU-hour; 1 LSU = $1 Runtime memory0.0090 LSU/GiB-hour Database compute and memory0.177 LSU/vCPU-hour plus 0.025 LSU/GiB-hour Free allowance / trial Plus includes one small serverless deployment; Developer has no Deployment access. Additional costs Tracing and model gateway usage can add charges; production deployments have database uptime costs. Watch for Do not merge this with LangSmith Sandboxes. The free Developer seat cannot deploy agents. Resource and session limits Isolation Managed agent server Persistence Deployment database persists state while live Regions LangSmith organization data region: US or EU GPU Not published Maximum session Not published Concurrency (running at once) Production autoscaling up to 10 containers by default Workload fit LangGraph agents needing managed checkpoints and API hosting. Source record PricingNov 3, 2026 Deployment data planeNov 3, 2026
	Letta Cloud and API PlanpublicSourceNov 3, 2026Hosts stateful agents and server-side tool execution in managed cloud environments.	$0Monthly starting price, not total	Personal Pro $20/month; API Plan $20/month plus $0.10/active agent-month and $0.00015/tool-execution second. Active agent$0.10/agent-month on API Plan Server-side tool execution$0.00015/second on API Plan	LLM tokens are billed at model rates unless BYOK is used; web search/fetch tools can add charges.	Personal Pro and API Plan serve different use cases; do not treat personal included quota as API credits.
Plans and sources for Letta / Letta Cloud and API Plan PRICE RECORD / letta-cloud Letta / Letta Cloud and API Plan Hosts stateful agents and server-side tool execution in managed cloud environments. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Personal Free $0/month Up to 3 stateful agents with BYOK; 5 MiB per memory repository Limited hosted usage Personal Pro $20/month plus pay-as-you-go overage Up to 20 agents and 50 MiB per memory repository Letta Auto quota; remote sandboxes available API Plan $20/month plus metered active agents, tool seconds and LLM tokens Unlimited agents and API key authentication $0.10/active agent-month; $0.00015/server-side tool second Teams Pro $20/seat/month Shared agents and permissions Usage quotas and overage apply Enterprise Custom Volume pricing, SSO and support Quote Usage meters Active agent$0.10/agent-month on API Plan Server-side tool execution$0.00015/second on API Plan Free allowance / trial Personal Free is recurring but restricted; no free API Plan was published. Additional costs LLM tokens are billed at model rates unless BYOK is used; web search/fetch tools can add charges. Watch for Personal Pro and API Plan serve different use cases; do not treat personal included quota as API credits. Resource and session limits Isolation Managed cloud sandbox for cloud agents Persistence Stateful agent memory and cloud session history Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) Not published Workload fit Stateful agents with hosted memory and API access. Source record Current pricingNov 3, 2026 Cloud agentsNov 3, 2026
	Mastra Platform ServerpublicSourceNov 3, 2026Deploys Mastra agents and workflows to a managed server.	$0Monthly starting price, not total	Starter $0 with 24 CPU-hours/month; Teams $250/month with 250 CPU-hours; compute overage and optional always-on server are extra. Persistent server$100/project/month on Starter and Teams Data egress$0.10/GB after plan allowance Model gatewayMarket token rate + 5.5%	Memory tokens, retrieval storage, database compute and storage, observability overage, egress and LLMs can bill separately.	The open-source Mastra framework is not the hosted Platform. Always-on service is an additional $100/project/month.
Plans and sources for Mastra / Mastra Platform Server PRICE RECORD / mastra-platform Mastra / Mastra Platform Server Deploys Mastra agents and workflows to a managed server. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Starter $0/month; then $0.35/CPU-hour 24 CPU-hours, 10 GB egress, 100,000 observability events/month 15-day retention Teams $250/month; then $0.25/CPU-hour 250 CPU-hours, 100 GB egress, 1 million observability events/month 6-month retention Enterprise Custom Custom volume, retention and controls Quote Usage meters Persistent server$100/project/month on Starter and Teams Data egress$0.10/GB after plan allowance Model gatewayMarket token rate + 5.5% Free allowance / trial Starter includes recurring 24 CPU-hours and 100,000 observability events monthly. Additional costs Memory tokens, retrieval storage, database compute and storage, observability overage, egress and LLMs can bill separately. Watch for The open-source Mastra framework is not the hosted Platform. Always-on service is an additional $100/project/month. Resource and session limits Isolation Managed server Persistence Optional persistent server; database and memory separately metered Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) Not published Workload fit Teams already building agents in Mastra. Source record PricingNov 3, 2026
	Copilot StudiopublicSourceNov 3, 2026Hosts low-code agents and agent flows for Microsoft and external channels.	$0Monthly starting price, not total	Pay as you go is $0.01/Copilot Credit with no up-front license; 25,000-credit capacity pack is $200/month; prepaid tiers are separate. Copilot Credit PAYG$0.01/credit Classic answer1 credit Generative answer2 credits Agent action5 credits	Graph grounding, voice, AI tools and flow actions use other credit amounts. Connected Azure resources and qualifying Microsoft 365 licenses have separate terms.	This is a low-code agent service, not a custom-code container runtime. A response can trigger several billable credit categories; Azure subscription required.
Plans and sources for Microsoft / Copilot Studio PRICE RECORD / microsoft-copilot-studio Microsoft / Copilot Studio Hosts low-code agents and agent flows for Microsoft and external channels. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Pay as you go $0.01/Copilot Credit in arrears; no up-front license commitment Hosted agents and agent flows Azure subscription and billing plan required Copilot Credit pack $200/pack/month 25,000 Copilot Credits per pack Tenant-wide license; credits consumed by actions and responses Pre-purchase One-year prepaid tiers; price varies by commitment Commit Units offset eligible usage Unused units do not roll over; compare official licensing guidance Usage meters Copilot Credit PAYG$0.01/credit Classic answer1 credit Generative answer2 credits Agent action5 credits Free allowance / trial Test-chat activity is not billed, but trial agents cannot be published; no recurring production credit allowance verified. Additional costs Graph grounding, voice, AI tools and flow actions use other credit amounts. Connected Azure resources and qualifying Microsoft 365 licenses have separate terms. Watch for This is a low-code agent service, not a custom-code container runtime. A response can trigger several billable credit categories; Azure subscription required. Resource and session limits Isolation Microsoft-managed low-code agent runtime Persistence Conversation and flow state depend on configuration Regions Microsoft availability and data residency vary by tenant GPU Not applicable to the low-code agent plan Maximum session Not published Concurrency (running at once) Tenant quota-dependent Workload fit Business agents that use Microsoft 365 and Power Platform. Source record Licensing guidanceNov 3, 2026 Plan pricingNov 3, 2026 Activity credit ratesNov 3, 2026
	Foundry Agent Service Hosted AgentsunresolvedSourceNov 3, 2026Runs custom agent code in Foundry-managed customer-dedicated containers.	Not publishedMonthly starting price, not total	Hosted container compute is billed per hour; a public USD unit rate was not exposed on the official pricing page. See plan details for usage rates.	Model tokens, Foundry Tools, Foundry IQ, connectors and licensed data are separate.	Do not use the no-additional-charge Foundry-native agent statement as the Hosted Agents compute price.
Plans and sources for Microsoft Azure / Foundry Agent Service Hosted Agents PRICE RECORD / azure-foundry-hosted-agents Microsoft Azure / Foundry Agent Service Hosted Agents Runs custom agent code in Foundry-managed customer-dedicated containers. Status: unresolvedChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Foundry-native agents No additional service charge; model and tool charges apply Prompt and workflow agents Not the custom-code Hosted Agents compute tier Hosted Agents Container compute per hour; public rate unavailable Managed custom-code containers and scaling Model tokens, tools, knowledge connections and licenses separate Usage meters Not separately published Free allowance / trial No Hosted Agents recurring free allowance verified. Additional costs Model tokens, Foundry Tools, Foundry IQ, connectors and licensed data are separate. Watch for Do not use the no-additional-charge Foundry-native agent statement as the Hosted Agents compute price. Resource and session limits Isolation Customer-dedicated managed containers Persistence Session-level state supported Regions Published region list in Hosted Agents documentation GPU Not published Maximum session Not published Concurrency (running at once) Managed scaling; numerical cap not verified Workload fit Custom-code agents that need Azure identity and governance. Source record Official pricingNov 3, 2026 Hosted Agents documentationNov 3, 2026
	n8n CloudpublicSourceNov 3, 2026Hosts workflow and agent automations with managed execution.	Not publishedMonthly starting price, not total	Starter EUR20/month and Pro EUR50/month when billed annually; USD price and monthly-billed rates were not verified. See plan details for usage rates.	Model usage may require external API keys; published Assistant credits are not workflow executions.	The Business tier is self-hosted only. EUR annual-billing display cannot be treated as a USD monthly-billed price.
Plans and sources for n8n / n8n Cloud PRICE RECORD / n8n-cloud n8n / n8n Cloud Hosts workflow and agent automations with managed execution. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Starter annual billing EUR20/month, billed annually 2,500 workflow executions/month; unlimited steps; agents and AI nodes 5 concurrent executions; 1 shared project Pro annual billing EUR50/month, billed annually 10,000 executions/month Up to 50 concurrent executions; 3 shared projects Business EUR667/month, billed annually 40,000 executions/month Self-hosted only; exclude from Cloud entry Enterprise Custom Hosted or self-hosted; custom executions Quote Usage meters Not separately published Free allowance / trial Hosted Starter has a free trial, not a permanent free plan. Additional costs Model usage may require external API keys; published Assistant credits are not workflow executions. Watch for The Business tier is self-hosted only. EUR annual-billing display cannot be treated as a USD monthly-billed price. Resource and session limits Isolation Managed n8n Cloud instance Persistence Managed workflow history; duration by plan not fully published Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) 5 Starter; up to 50 Pro; 200+ Enterprise Workload fit Hosted integrations and agent workflows priced by executions. Source record PricingNov 3, 2026
	Agent RunnerspublicSourceNov 3, 2026Runs coding agents against Netlify projects in a managed build environment.	$0Monthly starting price, not total	Uses Netlify plan credits for agent compute at 10 credits/GB-hour and model inference at 180 credits per $1 of model usage. Agent compute10 credits/GB-hour Agent Runner inference180 credits per $1 of model usage	Deploys, functions, bandwidth and other Netlify usage draw from the same plan credits.	This runner changes code in Netlify projects. It is not a general agent API runtime. Netlify hosting has a separate record.
Plans and sources for Netlify / Agent Runners PRICE RECORD / netlify-agent-runners Netlify / Agent Runners Runs coding agents against Netlify projects in a managed build environment. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Free $0/month; 300 shared credits/month limit 1 concurrent agent run Credits shared with other Netlify usage Personal $9-$15/month display varies; verify in account dashboard 1,000 shared credits/month; 3 concurrent agent runs Displayed plan price varies by billing context Pro $20-$25/month display varies; verify in account dashboard At least 3,000 shared credits/month; 10 concurrent agent runs Plan credit tier selection Enterprise Custom 50 concurrent agent runs listed Quote Usage meters Agent compute10 credits/GB-hour Agent Runner inference180 credits per $1 of model usage Free allowance / trial Free has a 300-credit monthly ceiling shared across Netlify usage; not 300 dedicated agent credits. Additional costs Deploys, functions, bandwidth and other Netlify usage draw from the same plan credits. Watch for This runner changes code in Netlify projects. It is not a general agent API runtime. Netlify hosting has a separate record. Resource and session limits Isolation Managed coding-agent run Persistence Project and deploy state; run filesystem persistence not published Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) 1 Free, 3 Personal, 10 Pro, 50 Enterprise Workload fit Code-change agents working inside Netlify projects. Source record Plan pricingNov 3, 2026 AI feature pricingNov 3, 2026
	Ona AgentsCustom quoteSourceNov 3, 2026Runs coding agents in isolated managed development environments.	Not publishedMonthly starting price, not total	Current pricing URL redirects to contact sales; current Core price, compute-unit conversion and plan tiers are not published there. See plan details for usage rates.	Ona compute units can cover environment runtime and agent conversations in older indexed material, but current conversion and model terms need confirmation.	Ona is a coding-agent environment, not a general production agent-serving API. Do not use cached Core $20 pricing after the live sales redirect.
Plans and sources for Ona / Ona Agents PRICE RECORD / ona-agents Ona / Ona Agents Runs coding agents in isolated managed development environments. Status: quoteChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Current hosted offering Quote; current self-serve rate not verified Isolated cloud agent environments and automations Current pricing URL redirects to sales; older indexed Core $20/month is not current evidence Usage meters Not separately published Free allowance / trial Current recurring free allowance not verified. Additional costs Ona compute units can cover environment runtime and agent conversations in older indexed material, but current conversion and model terms need confirmation. Watch for Ona is a coding-agent environment, not a general production agent-serving API. Do not use cached Core $20 pricing after the live sales redirect. Resource and session limits Isolation Isolated development environment per agent Persistence Environment lifecycle policy not verified on current price page Regions Not published GPU Not published on current price page Maximum session Not published Concurrency (running at once) Not published on current price page Workload fit Coding agents that need a managed development environment and pull-request workflow. Source record Current pricing redirectNov 3, 2026 Ona Agent documentationNov 3, 2026
	Generative AI ApplicationsunresolvedSourceNov 3, 2026Runs containerized agent code behind OCI-managed endpoints.	Not publishedMonthly starting price, not total	Managed agent deployment is documented; a current public USD rate for Generative AI Applications was not found. See plan details for usage rates.	Container Registry, model inference, storage and network usage may add costs; exact allocation to this product was not verified.	Oracle Generative AI model and Agents Service prices are not proof of Generative AI Applications runtime prices.
Plans and sources for Oracle Cloud Infrastructure / Generative AI Applications PRICE RECORD / oracle-genai-applications Oracle Cloud Infrastructure / Generative AI Applications Runs containerized agent code behind OCI-managed endpoints. Status: unresolvedChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Managed application Current hosted runtime rate unpublished in sources checked Container deployment, scaling and managed endpoint OCI compute, model and storage charges need a product-specific quote or rate card Usage meters Not separately published Free allowance / trial No product-specific recurring free allowance verified. Additional costs Container Registry, model inference, storage and network usage may add costs; exact allocation to this product was not verified. Watch for Oracle Generative AI model and Agents Service prices are not proof of Generative AI Applications runtime prices. Resource and session limits Isolation Managed containerized application Persistence Deployment configuration persists; request-state durability not published Regions OCI Generative AI availability varies by region GPU Not published for hosted applications Maximum session Not published Concurrency (running at once) Automatic scaling from zero to multiple instances; cap not verified Workload fit Containerized agents that need OCI identity and managed endpoints. Source record Managed agent deploymentNov 3, 2026 Application deployment modelNov 3, 2026 Container comparisonNov 3, 2026
	Pipedream WorkflowspublicSourceNov 3, 2026Runs production integration and agent workflows on Pipedream-managed infrastructure.	$150Monthly starting price, not total	Free development only; Startup $150/month billed monthly or $99/month billed annually, with 10,000 usage credits. Additional user$2/user after 100 included on Startup	Usage credits are a separate consumption pool; model or connected-service charges may apply.	Do not price production at $0, and do not label $99 annual-billed as month-to-month.
Plans and sources for Pipedream / Pipedream Workflows PRICE RECORD / pipedream-workflows Pipedream / Pipedream Workflows Runs production integration and agent workflows on Pipedream-managed infrastructure. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Free development $0 Development access to connectors and tools Not production execution Startup monthly $150/month Production workflows, 100 users and 10,000 usage credits Then $2 per additional user; usage beyond credits separate Startup annual $99/month, billed annually Same Startup tier Annual commitment Business Custom; annual billing Volume terms and enterprise controls Quote Usage meters Additional user$2/user after 100 included on Startup Free allowance / trial Free is for development; no recurring free production allowance verified. Additional costs Usage credits are a separate consumption pool; model or connected-service charges may apply. Watch for Do not price production at $0, and do not label $99 annual-billed as month-to-month. Resource and session limits Isolation Managed workflow execution Persistence Workflow state and logs; retention not published on pricing page Regions Not published GPU Not published Maximum session Not published Concurrency (running at once) Not published Workload fit Agent workflows that rely on managed SaaS connectors. Source record PricingNov 3, 2026
	Prefect Cloud and ServerlesspublicSourceNov 3, 2026Orchestrates workflows and includes Prefect-hosted Serverless execution minutes.	$0Monthly starting price, not total	Hobby free with 500 Serverless minutes; Starter $100/month with 75 hours; Team $100/user/month with 225 hours. See plan details for usage rates.	Bring-your-own-compute work pools run on customer infrastructure; Serverless overage rate not verified.	Prefect Cloud control-plane pricing and Prefect Serverless compute allowance are distinct; no public overage rate was verified.
Plans and sources for Prefect / Prefect Cloud and Serverless PRICE RECORD / prefect-cloud Prefect / Prefect Cloud and Serverless Orchestrates workflows and includes Prefect-hosted Serverless execution minutes. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Hobby $0 500 Serverless minutes/month; 5 deployments 2 users; 7-day run retention Starter $100/month 75 Serverless hours/month; 20 deployments 3 users; 1,250 API requests/minute Team $100/user/month 225 Serverless hours/month; 100 deployments 4-8 users listed; confirm seat minimum before budgeting Enterprise Custom Security and private-network options Quote Usage meters Not separately published Free allowance / trial 500 Prefect Serverless minutes recur monthly on Hobby. Additional costs Bring-your-own-compute work pools run on customer infrastructure; Serverless overage rate not verified. Watch for Prefect Cloud control-plane pricing and Prefect Serverless compute allowance are distinct; no public overage rate was verified. Resource and session limits Isolation Prefect Serverless or customer work pool Persistence Run retention 7 days on Hobby; other periods vary Regions Not published on pricing page GPU Not published for Serverless Maximum session Not published Concurrency (running at once) Not published Workload fit Python workflows with optional managed execution. Source record PricingNov 3, 2026
	Render WorkflowspublicSourceNov 3, 2026Runs durable task and flow code on Render Workflows compute plans.	Not publishedMonthly starting price, not total	Flex: $0.20/CPU-hour plus $0.05/GB-hour RAM used; fixed plans from $0.40 to $2.50/hour; state $0.25/GB. Flex CPU$0.20/cumulative virtual CPU-hour Flex RAM$0.05/GB-hour sampled usage Run-state retention$0.25/GB retained for 30 days; billed once per run/step state	Public internet outbound bandwidth contributes to workspace usage; the required workspace plan and its price are separate.	Flex uses measured CPU/RAM while fixed plans bill selected size. State and outbound bandwidth add cost; Render hosting plans are separate.
Plans and sources for Render / Render Workflows PRICE RECORD / render-workflows Render / Render Workflows Runs durable task and flow code on Render Workflows compute plans. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. flex $0.20/CPU-hour plus $0.05/GB-hour RAM used Up to 1 CPU and 4 GB RAM; per-second billing One-second minimum charge at 0.1 CPU and 0.1 GB RAM 2c-4g $0.40/hour 2 CPU, 4 GB RAM Per-second proration 2c-8g $0.70/hour 2 CPU, 8 GB RAM Per-second proration 4c-8g $1.00/hour 4 CPU, 8 GB RAM Per-second proration 4c-16g $1.50/hour 4 CPU, 16 GB RAM Per-second proration 8c-32g $2.50/hour 8 CPU, 32 GB RAM Per-second proration Usage meters Flex CPU$0.20/cumulative virtual CPU-hour Flex RAM$0.05/GB-hour sampled usage Run-state retention$0.25/GB retained for 30 days; billed once per run/step state Free allowance / trial No recurring free Workflows allowance stated on the Workflows limits page. Additional costs Public internet outbound bandwidth contributes to workspace usage; the required workspace plan and its price are separate. Watch for Flex uses measured CPU/RAM while fixed plans bill selected size. State and outbound bandwidth add cost; Render hosting plans are separate. Resource and session limits Isolation Managed workflow run Persistence Arguments and return values retained 30 days Regions Not published on limits page GPU Not published Maximum session 2-hour default attempt; configurable up to 24 hours Concurrency (running at once) Total 10,000 CPU/40,000 GB active allocation per service; new-compute rate depends on workspace plan Workload fit Durable tasks with a choice of measured flex or fixed-size compute. Source record Workflows limits and pricingNov 3, 2026 Render pricing cross-checkNov 3, 2026
	Restate CloudpublicSourceNov 3, 2026Runs a managed durable-execution service while application handlers use separate compute.	$0Monthly starting price, not total	Free 100,000 actions/month; Starter $75 for 5 million; Business $300 for 20 million; Premium $1,000 for 50 million. Added Virtual Object storage$0.50/GB-month	Customer-hosted handler compute and model tokens are not included; optional compliance and multi-AZ add-ons cost extra.	Restate Cloud coordinates work but does not host agent code. Each function call counts as at least two actions.
Plans and sources for Restate / Restate Cloud PRICE RECORD / restate-cloud Restate / Restate Cloud Runs a managed durable-execution service while application handlers use separate compute. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Free $0/month 100,000 actions and 1 GB Virtual Object storage 10 actions/sec; 1-day history Starter $75/month; $25/million extra actions 5 million actions; 5 GB storage 50 actions/sec; 3-day history Business $300/month; $25/million extra actions 20 million actions; 20 GB storage 200 actions/sec; 7-day history Premium $1,000/month; $15/million extra actions 50 million actions; 50 GB storage 500 actions/sec; 30-day history Enterprise / BYOC Custom Custom capacity and deployment Quote Usage meters Added Virtual Object storage$0.50/GB-month Free allowance / trial 100,000 actions and 1 GB storage recur each month on Free. Additional costs Customer-hosted handler compute and model tokens are not included; optional compliance and multi-AZ add-ons cost extra. Watch for Restate Cloud coordinates work but does not host agent code. Each function call counts as at least two actions. Resource and session limits Isolation Managed orchestration service; handler isolation depends on customer compute Persistence Durable state with tiered history retention Regions Not published on pricing page GPU Not applicable to managed service Maximum session Not published Concurrency (running at once) Action throughput by tier, not agent concurrency Workload fit Durable coordination when the team already hosts its handlers. Source record PricingNov 3, 2026
	AgentforcepublicSourceNov 3, 2026Runs configured customer and employee agents inside Salesforce workflows.	Not publishedMonthly starting price, not total	Flex Credits $500 per 100,000-credit pack; standard actions use 20 credits ($0.10); conversation model $2/conversation; user license $5/user/month requires Flex Credits. Standard agent action20 Flex Credits = $0.10 at pack rate Agentforce Voice action30 Flex Credits	Salesforce CRM licenses, Data Cloud, connected services and other action types can change the total bill.	This is an enterprise application agent service, not a deploy-any-code runtime. Flex Credits are bought in packs, not a $500 monthly subscription.
Plans and sources for Salesforce / Agentforce PRICE RECORD / salesforce-agentforce Salesforce / Agentforce Runs configured customer and employee agents inside Salesforce workflows. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Salesforce Foundations $0 add-on shown for eligible Salesforce customers Agentforce Builder and related tools Underlying Salesforce edition and agent usage rights are separate Flex Credits $500 per 100,000 credits Customer- and employee-facing agent actions Standard action 20 credits; Voice action 30 credits Conversations $2/conversation Conversation-based purchase option Eligibility and overage terms depend on Salesforce agreement Agentforce User License $5/user/month plus required Flex Credits Employee access to Agentforce Salesforce CRM license prerequisites may apply Agentforce Max Editions From $550/user/month Agentforce add-on and annual credits Specific Salesforce edition and contract required Usage meters Standard agent action20 Flex Credits = $0.10 at pack rate Agentforce Voice action30 Flex Credits Free allowance / trial Foundations lists a $0 add-on for eligible customers; the free-credit amount depends on Salesforce edition and is not a universal runtime allowance. Additional costs Salesforce CRM licenses, Data Cloud, connected services and other action types can change the total bill. Watch for This is an enterprise application agent service, not a deploy-any-code runtime. Flex Credits are bought in packs, not a $500 monthly subscription. Resource and session limits Isolation Salesforce-managed low-code agent platform Persistence CRM records and workflow state inside Salesforce Regions Salesforce deployment region depends on customer org GPU Not applicable Maximum session Not published Concurrency (running at once) Not published Workload fit Agents that act on Salesforce data and flows. Source record Current Agentforce pricingNov 3, 2026 Action-price explanationNov 3, 2026
	Cortex AgentspublicSourceNov 3, 2026Runs managed data agents and tools in a Snowflake account.	Not publishedMonthly starting price, not total	Agent orchestration bills AI Credits per million model tokens; tool and warehouse usage add separate charges. USD per credit is contract-dependent. OrchestrationAI Credits per million tokens by selected model	Cortex Search, Cortex Analyst, warehouse SQL, custom tools, code execution and web search can add costs.	This is a managed data-agent platform, not a general customer-code container runtime. AI Credit to USD conversion and tool costs vary by account.
Plans and sources for Snowflake / Cortex Agents PRICE RECORD / snowflake-cortex-agents Snowflake / Cortex Agents Runs managed data agents and tools in a Snowflake account. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Cortex Agents consumption AI Credits per million tokens; USD rate varies by model and Snowflake credit agreement Managed agent orchestration and REST API No flat agent-hosting subscription rate published Usage meters OrchestrationAI Credits per million tokens by selected model Free allowance / trial No universal recurring Cortex Agents allowance verified. Additional costs Cortex Search, Cortex Analyst, warehouse SQL, custom tools, code execution and web search can add costs. Watch for This is a managed data-agent platform, not a general customer-code container runtime. AI Credit to USD conversion and tool costs vary by account. Resource and session limits Isolation Snowflake-managed agent and optional isolated code execution Persistence Agent objects persist in Snowflake; per-run state depends on tools Regions Model and cross-region inference availability vary by Snowflake region GPU Not published for Cortex Agents Maximum session Not published Concurrency (running at once) Account and service quota-dependent Workload fit Governed agents over Snowflake data. Source record Cortex Agents productNov 3, 2026 Snowflake AI pricingNov 3, 2026
	Temporal CloudpublicSourceNov 3, 2026Hosts durable workflow orchestration while application workers run on customer compute.	$0Monthly starting price, not total	No base minimum; actions start at $50/million, with volume rates as low as $25/million; active and retained storage extra. Active storage$0.042/GB-hour Retained storage$0.00105/GB-hour	Workers run on customer infrastructure and cost extra; support options can add a minimum.	Temporal coordinates workflows but does not run your worker code. Its action charges are not comparable to hosted CPU prices.
Plans and sources for Temporal / Temporal Cloud PRICE RECORD / temporal-cloud Temporal / Temporal Cloud Hosts durable workflow orchestration while application workers run on customer compute. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Pay as you go $0 base; from $50/million actions No minimum; 99.9% SLA Storage and worker compute separate Volume discounts As low as $25/million actions self-service; further discounts by commitment Higher-volume action pricing Volume-dependent Business support Greater of $500/month or 10% of usage Support plus 2.5 million actions, 2.5 GB active and 100 GB retained storage Support add-on, not required base plan Enterprise / Mission Critical Custom Advanced support and commitments Quote Usage meters Active storage$0.042/GB-hour Retained storage$0.00105/GB-hour Free allowance / trial $150 in credits for 90 days is one-time, not monthly. Additional costs Workers run on customer infrastructure and cost extra; support options can add a minimum. Watch for Temporal coordinates workflows but does not run your worker code. Its action charges are not comparable to hosted CPU prices. Resource and session limits Isolation Managed orchestration; customer-hosted workers Persistence Durable workflow history and retained storage Regions Not published on pricing page GPU Worker-dependent Maximum session Not published Concurrency (running at once) Not published Workload fit Long-lived durable workflows with separately managed workers. Source record PricingNov 3, 2026
	Trigger.dev CloudpublicSourceNov 3, 2026Runs background tasks and agent code on managed workers.	$0Monthly starting price, not total	Free $5 monthly usage credit; Hobby $10/month with $10 credit; Pro $50/month with $50 credit; per-second compute plus $0.000025 per run. Small 1x$0.0000338/second (0.5 vCPU, 0.5 GB) Medium 1x$0.0000850/second (1 vCPU, 2 GB) Run invocation$0.000025/run in production	Compute and invocation both apply; BAA, extra seats, schedules, branches, dashboards and Realtime capacity can add charges.	Plan credits offset usage; they are not a second fee. Every started production run has an invocation charge.
Plans and sources for Trigger.dev / Trigger.dev Cloud PRICE RECORD / trigger-cloud Trigger.dev / Trigger.dev Cloud Runs background tasks and agent code on managed workers. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Free $0/month plus usage after $5 monthly credit 20 concurrent runs, 10 schedules, 1-day logs 5 team members Hobby $10/month with $10 monthly credit 50 concurrent runs, 100 schedules, 7-day logs 5 team members Pro $50/month with $50 monthly credit 200 concurrent runs, 1,000 schedules, 30-day logs Extra concurrency $10/month per 50 runs; other add-ons Enterprise Custom Custom discounts and controls Quote Usage meters Small 1x$0.0000338/second (0.5 vCPU, 0.5 GB) Medium 1x$0.0000850/second (1 vCPU, 2 GB) Run invocation$0.000025/run in production Free allowance / trial Recurring $5 usage credit on Free, not a $5 cash payout. Additional costs Compute and invocation both apply; BAA, extra seats, schedules, branches, dashboards and Realtime capacity can add charges. Watch for Plan credits offset usage; they are not a second fee. Every started production run has an invocation charge. Resource and session limits Isolation Managed task worker Persistence Durable waits checkpoint and do not bill compute while waiting Regions Not published GPU Not published Maximum session No fixed maximum run duration Concurrency (running at once) 20 Free, 50 Hobby, 200 Pro concurrent runs Workload fit Long-running JavaScript tasks and agents needing managed execution. Source record PricingNov 3, 2026
	Vercel WorkflowspublicSourceNov 3, 2026Runs durable workflows on Vercel with separately billed function compute and workflow state.	$0Monthly starting price, not total	50,000 workflow events/month included; then $20/million events, $0.50/GB writes and $0.50/GB-month retained state on Pro. Workflow events$20/million after 50,000/month Workflow storage writes$0.50/GB after 1 GB Workflow state retention$0.50/GB-month	Vercel Functions, Queues, AI Gateway and model tokens use separate meters.	Do not treat the Workflow SDK as a separate hosted agent server. Older docs list steps and GB-hours, while current pricing lists events and GB writes/retention; verify before quoting a bill.
Plans and sources for Vercel / Vercel Workflows PRICE RECORD / vercel-workflows Vercel / Vercel Workflows Runs durable workflows on Vercel with separately billed function compute and workflow state. Status: publicChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Hobby $0/month hosting plan; included Workflow events shown 50,000 workflow events/month Usage after Hobby limits may require Pro; check account terms Pro General hosting plan charge applies; Workflow usage metered separately 50,000 Workflow events/month and 1 GB storage writes Function compute and Queues bill separately Enterprise Custom Contract workflow terms Quote Usage meters Workflow events$20/million after 50,000/month Workflow storage writes$0.50/GB after 1 GB Workflow state retention$0.50/GB-month Free allowance / trial 50,000 Workflow events/month on the current pricing page; not a complete free agent compute allowance. Additional costs Vercel Functions, Queues, AI Gateway and model tokens use separate meters. Watch for Do not treat the Workflow SDK as a separate hosted agent server. Older docs list steps and GB-hours, while current pricing lists events and GB writes/retention; verify before quoting a bill. Resource and session limits Isolation Vercel Function steps with durable orchestration Persistence Billed workflow state retention Regions Depends on Vercel compute region GPU Not published Maximum session Durable workflow can wait beyond one function execution Concurrency (running at once) Not published Workload fit Durable agents already running on Vercel Functions. Source record Current pricingNov 3, 2026 Workflow productNov 3, 2026
	Windmill CloudunresolvedSourceNov 3, 2026Hosts scripts, workflows and agent jobs on Windmill-managed workers.	Not publishedMonthly starting price, not total	Community, Team and Enterprise Cloud plans are documented, but their current Cloud dollar rates were not verifiable from the accessible pricing view. See plan details for usage rates.	Cloud Team bills seats and extra executions; exact rates are not verified. Model-provider calls can add separate cost.	The accessible pricing calculator showed self-hosted worker and seat rates. Do not use them as Windmill Cloud rates.
Plans and sources for Windmill / Windmill Cloud PRICE RECORD / windmill-cloud Windmill / Windmill Cloud Hosts scripts, workflows and agent jobs on Windmill-managed workers. Status: unresolvedChecked: Nov 3, 2026 Published plans Other plans and regional prices may be available at the source. Community Cloud Public rate not verified Hosted scripts and workflows Cloud allowance and concurrency not verified Team Cloud Usage-based; public rate not verified Paid workspace seats and executions One seat includes 10,000 executions according to Windmill terms Enterprise Cloud Custom Dedicated hosted instance and support Quote Usage meters Not separately published Free allowance / trial Community Cloud allowance not verified; self-hosted Community is a separate product. Additional costs Cloud Team bills seats and extra executions; exact rates are not verified. Model-provider calls can add separate cost. Watch for The accessible pricing calculator showed self-hosted worker and seat rates. Do not use them as Windmill Cloud rates. Resource and session limits Isolation Managed Windmill workers Persistence Workflow run history; retention varies by plan Regions Not verified GPU Not published Maximum session Not published Concurrency (running at once) Not verified Workload fit Teams using Windmill scripts and workflows after Cloud pricing confirmation. Source record Pricing pageNov 3, 2026 Cloud plan documentationNov 3, 2026 Cloud Team billing termsNov 3, 2026

Prices in USD unless noted. Products start in alphabetical order. Compute estimates exclude extras and do not measure performance.

METHODOLOGY

What the prices include

Sources and dates

Prices come from provider pages, not live quotes. Each product has source links and a check date. This index does not cover every cloud plan. An unknown price is not zero.

Fees and usage

The entry price is a monthly starting price, not the full bill. Some plans include usage credit. Compute estimates include the selected plan and compute usage. The storage and outbound estimate adds those costs where rates are known. Taxes, model APIs, and other extras remain separate. One-time trials are not monthly credits.

Resources and performance

Equal CPU and memory counts do not mean equal performance. Shared and dedicated CPUs differ. GB and GiB are different units; selected rates show the source unit. Check the region, limits, and current terms at the source.

JSON data

The three parts of the bill

Not published / month starting price

Usage charges

Models & other costs

Compare products

Anthropic / Claude Managed Agents

Published plans

Managed Agents beta

Usage meters

Free allowance / trial

Additional costs

Watch for

Resource and session limits

Workload fit

Source record

AWS / Amazon Bedrock AgentCore Runtime

Published plans

microVM v1

microVM v2

Instances

Usage meters

Free allowance / trial

Additional costs

Watch for

Resource and session limits

Workload fit

Source record

Cloudflare / Workers and Agents

Published plans

Workers Free

Workers Paid

Usage meters

Free allowance / trial

Additional costs

Watch for

Resource and session limits

Workload fit

Source record

CrewAI / CrewAI Cloud

Published plans

Basic

Enterprise

Usage meters

Free allowance / trial

Additional costs

Watch for

Resource and session limits

Workload fit

Source record

Databricks / Agent Bricks Agent Runtime

Published plans

Free Edition evaluation

Medium App

Large App

XLarge App

Usage meters

Free allowance / trial

Additional costs

Watch for

Resource and session limits

Workload fit

Source record

DBOS / DBOS Cloud

Published plans

DBOS Cloud

Usage meters

Free allowance / trial

Additional costs

Watch for

Resource and session limits

Workload fit

Source record

Dify / Dify Cloud

Published plans

Sandbox

Professional monthly

Professional annual

Team monthly

Team annual

Enterprise