Skip to main content
This page is written for AI agents (and the developers building them). If you’re an agent reading this to implement SentFromAI, you have everything you need below.
Shortest path — paste this into your agent and it sets everything up itself:
The guide tells the agent how to reuse your account, save credentials securely, and return a private claim link only when needed. Set up your agent.

​
Ingest the whole docs

The full documentation is available as a single plaintext file optimized for LLMs:
There’s also an index at /llms.txt. Append .md to any docs page URL to get its raw Markdown.

​
The essentials

  • Base URL: https://api.sentfrom.ai/v1
  • Auth: Authorization: Bearer sf_live_… on authenticated requests. POST /agent-signups bootstraps a key without authentication when enabled (details).
  • Format: JSON in, JSON out. Errors are { "error": "…" } with standard status codes (reference).

​
Start with an account

Follow Set up your agent: reuse a saved key or bootstrap an inbox through REST, then hand the private claim link to your operator. Unclaimed accounts can receive and read within their limits; sending and other changes require a human claim. Read current capabilities through GET /account.

​
The core loop

After claiming the account, agents can use this loop. Reuse the signup inbox; create more only when needed and permitted by account limits:
1

Create an inbox

POST /inboxes { local_part } → an address like agent@mail.sentfrom.ai, live immediately. (Inboxes)
2

Send mail

POST /messages { inbox_id, to, subject, text }. Use client_id for idempotent retries. (Sending)
3

Receive mail

Register a webhook for message.received, or stream events over the realtime WebSocket, or poll GET /messages.
4

Reply in thread

POST /messages/{id}/reply { text } — threading headers are set for you. (Threads)

​
Use it as native tools (MCP)

Rather than calling the REST API, connect the SentFromAI MCP server — sentfromai-mcp on npm — and all of the above becomes 21 tools the model calls directly:
Running OpenClaw, LangChain, CrewAI, or another framework? There’s a per-framework recipe for each.

​
Worked examples

Auto-reply agent

Webhook → generate a reply → send it in thread.

Email labeling agent

Classify inbound mail and apply labels.
Reuse the key from agent setup, or create one manually in the dashboard → API keys. After claiming, the free plan gives 5 inboxes, 5,000 emails/month, and 5 GB of attachment storage — enough to run a real agent, not just a demo. Paid plans start at $19/mo for 100 inboxes and 20,000 emails.