Ship APIs, workflows & AI agents you can actually see.
Build from a prompt or by hand, and go live in seconds. No black boxes — every endpoint is observable the moment it ships. Cron schedules · Flow control · Durable state.
Built like a backend. Governed like a platform.
Air Pipe turns a single config into a running API — with the platform features you'd otherwise wire up yourself. Managed or self-hosted.
A real backend that runs your workflows — Cron schedules, branching and retries live in the same config as your HTTP APIs — with your database behind them and auth in front. One artifact, reviewable in a pull request.
Observable by default — Every endpoint ships OpenTelemetry traces, Prometheus metrics and OpenAPI docs, generated automatically. No black boxes, no separate instrumentation.
Your data, your infrastructure — Run on our managed cloud or self-host the single binary anywhere. Your data and your embeddings never leave your infrastructure.
Built for the people who own the backend
Air Pipe is a developer platform for APIs, workflows and AI agents. It is not air-quality monitoring, HVAC, or physical pipe infrastructure.
Backend & platform engineers — Ship services without standing up another framework, ORM layer and instrumentation stack for every internal tool.
Teams shipping B2B APIs — Multi-tenant APIs with auth, plan limits and per-customer isolation — defined as config you can read in a PR.
AI agent & MCP builders — Expose your real systems as MCP tools agents can call, and see every call an agent made in a trace.
Self-host & compliance teams — Run the single binary in your own environment. Your data and your embeddings stay on your infrastructure.
Build it in minutes
Describe what you want in plain English and Air Pipe writes the config. Or write the YAML yourself, or fork a pack from the marketplace. There's no project to scaffold and no boilerplate to maintain.
Build with AI – describe it in a prompt, get a config validated against the live schema
Readable YAML – it's yours, and you can edit it by hand whenever you like
Marketplace packs – fork a working starter and deploy it
Connect your whole stack
Query your database, call any HTTP API, run a command, send email, or reach Google and AWS — then chain them together in a single config.
Databases – Postgres, MySQL, MongoDB and SQLite, wherever they're hosted
HTTP, commands & email – call upstream APIs, run scripts, send over SMTP
Google & AWS – cloud services as actions in your flow
Workflows and automation, built in
Schedules, webhooks and multi-step flows are the same config as your APIs — not a separate product with separate pricing.
Run on a schedule – or trigger it from a webhook
Flow control – conditions, transforms, asserts and state between steps
Every run traced – see what actually happened, not just that it failed
Secure by default
Authentication, access control and validation are part of the config, not something you bolt on afterwards.
Auth – JWT validation, API keys and role-based access control
Validation – assert on any input before it reaches your data
Network policy – control what each interface is allowed to reach
Fast, and yours to run
Built in Rust and shipped as a single binary. Run it on our cloud or your own infrastructure — the same config either way.
Managed or self-hosted – VM, container, or let us host it
Single binary – no sidecars, no agents to babysit
Priced per request – not per task, per execution, or per AI credit
Fork a pack. Make it yours.
Every pack is a real config you can deploy, read and edit — schedules and flows alongside APIs and agent tools. Start from one instead of a blank file.
Flow Control — Flow-control recipes for Air Pipe pipelines: throttle with the delay action, retry with exponential backoff, branch with run_on_assertion, try/catch/finally recovery, and at_least quorum. Every endpoint is self-contained — no external services, databases, or managed variables required.
Durable API Polling — Poll an API on a schedule and process only new items, retrying only the items that fail — never the whole batch and never silently dropping a failure. Durable id-based dedupe via the state feature plus lookup_partition per-item retry. Includes a self-contained demo and a production poller template.
Daily Database Digest Email — Query your database, format a summary email, and send it via Resend. Ships with a sample app_events table so you can try it immediately. Swap the query for any SQL that returns numbers your team cares about — signups, revenue, errors, API calls.
Stripe Webhooks → DB → Slack — Receives Stripe webhook events, verifies the signature, stores every event in Postgres with idempotency, and posts a Slack notification. One endpoint handles all event types.
REST API Starter — A production-shaped CRUD API over Postgres with three resource types (users, posts, comments) and optional JWT validation. No external services required — just a Postgres database. Deploy and start making requests in under 10 minutes.
Multi-Tenant SaaS Starter — A production-shaped multi-tenant backend over Postgres: API-key authentication (sha256-hashed, generated server-side, revocable), strict per-tenant data isolation, plan-based limits, and usage metering. Every request resolves its API key to exactly one tenant and every query is scoped to it — cross-tenant access is structurally impossible.
RAG Chatbot — Retrieval-Augmented Generation chatbot backed by Postgres + pgvector. Ingest your own documents, store OpenAI embeddings alongside them, and answer questions using only your data — no third-party vector DB, no proprietary cloud, nothing leaves your infrastructure. Includes seed, ingest, chat, history, and delete endpoints.
MCP Postgres Starter — Turn your Postgres data into secure MCP tools any AI client (Claude Desktop, Claude Code, Cursor) can call. Two-tier auth: one shared token for yourself, or per-user tenant-scoped tokens for your customers - minted via an HS256 exchange or verified directly against your OIDC provider's JWKS (Auth0, Clerk, Cognito). Every tool reuses the same in-config token checks as a normal Air Pipe route.