llms.txt
Compare with other sites' llms.txt files →# Postfleet
> Email infrastructure for AI agents. Inbound mail is parsed, classified, and
> screened for prompt injection before your agent reads it; outbound send is
> idempotent with optional human-in-the-loop approval. Every API endpoint lives
> under the canonical /api/v1/ prefix (base URL https://api.postfleet.ai).
## Docs
- [Quickstart](/docs/quickstart): Get an API key, send your first email, receive inbound mail via webhook, and connect over MCP.
- [Authentication & key scoping](/docs/auth): Bearer pf_ keys, mailbox-bound scoping, the data-plane vs control-plane split, and anti-enumeration 404s.
- [Sending & idempotency](/docs/sending): POST /api/v1/send, the single-recipient rule, client_id duplicate protection with the full replay table, and provider-unknown semantics.
- [Drafts & human-in-the-loop](/docs/drafts): Voluntary vs. forced drafts, the pending_approval flow, content_hash concurrency at approval, tombstones, and the MCP draft tools.
- [Send lists](/docs/lists): Allow/blocklist precedence, address normalization, exact-label matching, and the 403 send_blocked_by_list shape.
- [Webhooks](/docs/webhooks): The message.received event, event_id dedup, the custom HMAC signature recipe with a copy-pasteable verifier, at-least-once retries, and dead-letter redrive.
- [Comprehend (BYO inbox)](/docs/comprehend): Screen email you already have with POST /api/v1/comprehend — sanitization, injection scan, and optional extraction without provisioning a mailbox.
- [MCP setup](/docs/mcp): Configure the Postfleet MCP server with a bearer key, the full tool list, and how to handle the 202 pending-approval outcome.
- [TypeScript SDK](/docs/sdk): Install the Postfleet TypeScript SDK, send and wait for email from application code, and handle errors and idempotency.
- [API reference](/docs/api): The full /api/v1 REST surface — every endpoint, grouped by resource, with a link to the machine-readable OpenAPI description.
## Notes
- Every page above is fetchable as verbatim markdown at /docs/<slug>/raw.
- All examples use the canonical /api/v1/ prefix and bearer pf_ keys.
- /llms-full.txt inlines every page above in full, for one-fetch reading.
## Comparisons
- [Postfleet vs AgentMail](/compare): feature-by-feature comparison of two agent-inbox products — what each screens before an agent reads a message.
- [Postfleet vs SendGrid Inbound Parse](/compare/sendgrid-inbound-parse): Compare Postfleet with the Twilio SendGrid Inbound Parse Webhook for AI agents: what each does to an inbound email before your code or your model sees it, retries, storage, spam handling, and pricing.
- [Postfleet vs Gmail MCP Server](/compare/gmail-mcp-server): Compare a dedicated, screened agent inbox with giving an AI agent Gmail through OAuth scopes, the Gmail API, or a Gmail MCP server: what is built in, what you assemble, and what reaches the model.
- [Postfleet vs inbound.new](/compare/inbound-new): Compare Postfleet and inbound.new, two hosted inbound-email APIs for AI agents: webhooks, threading, MCP, the Inbound Guard rule engine versus enforced screening, authentication verdicts, pricing, and open source.
- [Postfleet vs LobsterMail](/compare/lobstermail): Compare Postfleet and LobsterMail, two inboxes for AI agents that screen for prompt injection: what safeBodyForLLM and the injection risk score do, what fail-closed means, MCP, webhooks, extraction, and pricing.
## Security
- [Security overview](/security): the living red-team scoreboard — measured catch rate, what got through, and the limits we don't claim past.
- [Prompt injection attack examples](/security/prompt-injection-examples): six real payload shapes (hidden HTML, comments, zero-width text, multi-step decode SOPs) with the actual sanitizer and scanner output for each.
- [Email prompt injection](/security/prompt-injection-email): the channel-level threat model — headers, HTML, the text/plain alternative, quoted history, attachments — plus the gate order and the five terminal comprehension statuses.