
Figment and Claw Messenger both give developers programmatic access to iMessage — but through completely different account models. Claw provisions its own Apple-registered lines and lets an AI agent send messages without touching the user's Apple Account. Figment connects the end-user's own Apple Account and gives your app scoped OAuth access to their real iMessage history and sender identity. The right choice comes down to one question: does your app need to operate from and read the user's actual Apple ID, or does it need an independent agent line that your user never has to authorize?
Claw Messenger is a managed iMessage, RCS, and SMS API aimed specifically at AI agents. The service provisions its own Apple-registered numbers. Your agent sends messages through those provider-owned lines, not through your user's personal Apple ID. Individual plans start at $5/month for 250 messages and scale to $50/month for 15,000 messages on shared numbers. An Agency tier is priced at $199/month per sub-tenant, includes 1,000 messages, and provides dedicated numbers with a $250 one-time parent-account onboarding fee (clawmessenger.com).
No Mac is required on your end. Claw runs on Linux, Docker, Windows, or cloud servers. Incoming replies stream over an authenticated WebSocket, so you need no public-facing webhook URL. The service ships SDKs and integration guides for Python, Node.js, LangChain, CrewAI, and its own OpenClaw library. Built-in SMS and RCS fallback means Android contacts are reachable without a separate carrier arrangement (clawmessenger.com/imessage-api).
The key constraint is identity. Recipients see a Claw-provisioned number, not your user's phone number or Apple ID email. Incoming message access covers only messages addressed to those provisioned lines — not the existing conversation threads in a user's Messages app.
Figment is an OAuth gateway to the end-user's own Apple Account. Your user connects through Figment's hosted consent screen — this is Figment's authorization flow; Figment issues the grant, code, and access token. Your app receives a scoped access token and never receives the Apple password. A dedicated signed-in Mac, hosted and managed by Figment, executes native Apple operations.
There are exactly two iMessage scopes. messages:read covers both conversation history and search. messages:send covers sending into existing conversations. Before reading history, check the historyComplete and hasMore flags — native history may be incomplete while the Mac is still syncing, and reads against a non-ready connector are rejected. SMS, MMS, and RCS messages are also available when the user has an eligible iPhone that is on, online, and signed into the same Apple Account. When Messages in iCloud is enabled, carrier messages sync without a separate forwarding toggle (support.apple.com/en-us/102545).
Pricing is $29/month per active live Apple connection. Sandbox access is free and uses synthetic data. Access starts after setup; immediate provisioning is not guaranteed.
This is the sharpest difference between the two services, and it determines almost every downstream design decision.
With Claw, the sender identity belongs to Claw's infrastructure. Your users do not connect their Apple Account; they give your application permission to use a Claw-managed line. Recipients see a Claw-provisioned number. Your app cannot read threads from the user's existing iMessage history — only messages exchanged through those Claw lines.
With Figment, the sender identity is the user's real Apple ID. Recipients see the user's actual phone number or Apple ID email address. Your app can read conversation history that already exists in the user's Messages app, including threads that predate the connection. The user grants access through Figment's consent screen and can revoke it at any time from /apple-connect/account.
One boundary worth noting: Apple's documented third-party app authorization for iCloud covers Mail, Calendar, and Contacts (support.apple.com/en-us/121539). Figment's iMessage access uses a Mac-based bridge that is distinct from that cloud-authorization path.
Claw Messenger is the better fit when:
Figment is the better fit when:
One boundary to plan for in either direction: Figment's messages:send grant requires the target conversation to already exist in the user's Messages app and be explicitly included in the send grant — a conversation visible in read results is not automatically authorized for sends. Cold outreach to a contact with no prior thread on that Apple Account requires that thread to exist and be granted first. Claw's model allows agents to initiate new contacts within published daily limits, without that constraint. Neither service guarantees delivery; always poll a receipt before assuming a message reached the recipient.
Facts below are drawn from each service's public documentation as of November 2026.
| Dimension | Claw Messenger | Figment |
|---|---|---|
| Sender identity | Provider-provisioned number | User's own Apple ID |
| Incoming message access | Messages to Claw's lines only | User's real iMessage history |
| Initiate new conversations | Yes (within published daily limits) | Existing granted conversations only |
| SMS / RCS | Built in | Requires eligible iPhone online (Messages in iCloud) |
| Mac on your infrastructure | Not required | Figment-hosted |
| Auth model | Claw API key | Figment OAuth with PKCE |
| Starting price | $5 / mo (250 messages) | $29 / mo per live connection |
| Free / trial tier | 7-day free trial | Free sandbox with synthetic data |
| Primary SDK targets | Python, Node, LangChain, CrewAI | Any HTTP client |
| Apple account owned by | Claw (provider) | End-user |
If your use case matches the user-identity model, you can test Figment's iMessage API against synthetic conversation data at no cost before connecting a live account. Register your app and generate credentials at /apple-connect/developers. When you're ready to run against a real Apple Account, users connect through the Connect Apple Account flow on Figment's consent screen. The per-connection price for a live account is $29/month.
Related reading: