FastClaw
Feature · API

An OpenAI-compatible API for your agents

Drive FastClaw agents from any app: OpenAI-compatible chat completions, agent management, per-end-user isolation, usage reporting and ACP.

Get started freeSee pricing
01

Drop-in chat completions

/v1/chat/completions speaks the OpenAI format, streams the agent’s text as it is written and reports each tool call as it runs.

02

Per-end-user isolation

Pass X-Fastclaw-End-User (or user) and FastClaw keeps sessions, USER.md and memory separate for every user of your app.

03

Manage agents over HTTP

Create, list, update and delete agents on /v1/agents and write their system files — with metadata filters and pagination.

04

Real usage numbers

Every turn reports real token counts, tool loops included, with cache reads and writes split out. Export usage events for billing.

05

Scoped API keys

admin keys for automation, user keys for multi-tenant apps, agent keys locked to an explicit list of agents.

06

ACP 0.2

An Agent Communication Protocol endpoint at /acp for agent discovery, manifests, sync, async and streamed runs.

Call an agent like a model

curl https://your-fastclaw/v1/chat/completions \
  -H "Authorization: Bearer $FASTCLAW_API_KEY" \
  -H "X-Fastclaw-End-User: user_123" \
  -d '{"agent_id": "agt_...", "stream": true, "messages": [{"role": "user", "content": "Hi"}]}'

The request runs the same agent loop as the web chat: skills, sandboxed tools, memory and all. Messages can carry images in OpenAI or Anthropic shape, files the agent creates come back with a download URL, and failures return a proper error — 502 agent_error — instead of a polite apology. Try the API snippet generator to get a ready-to-paste client.

Built for multi-tenant apps

A user key plus the end-user header turns one FastClaw into the agent backend for a whole product. Each of your users gets isolated sessions, memory and files without you managing FastClaw accounts. Admins can provision users and agents over the API, fork an agent's identity into a new account and cap self-service creation with an agent quota. This is the core of FastClaw for SaaS builders and agencies.

Hooks for your billing

FastClaw never charges anyone, but it gives you the pieces: per-turn usage on every response, a cursor-based usage event export with the paying account on each row, a billing hold that pauses an account (402 payment_required), and single-use login links into the console. State lives in SQLite or Postgres, so you can scale out behind a load balancer. Coding agents can wire it all up from the integration guide every FastClaw serves at /skills/agent-integration/SKILL.md.

Works with

All Integrations

OpenAI

Run FastClaw agents on OpenAI models — set a deployment-wide default, override per agent, and expose your agents through an OpenAI-compatible API.

PostgreSQL

Swap FastClaw's default SQLite file for PostgreSQL to run multiple pods behind a load balancer, with automatic schema migrations.

Frequently asked questions

More Features

All Features

Agents

Create and run many AI agents from one FastClaw install — each with its own SOUL.md personality, memory, skills, model and sharing settings.

Models

Run FastClaw agents on OpenAI, Anthropic, DeepSeek, Ollama, OpenRouter, Groq, Mistral or any OpenAI-compatible API, with per-agent overrides and retries.

Channels

Bind any FastClaw agent to Telegram, Discord, Slack, Feishu, WeCom, WeChat, WhatsApp, LINE or iMessage — with per-chat session isolation and owner pairing.

Try it on your own agents

Start free in FastClaw Cloud, or self-host the single binary in a minute.

Get started freeSee pricing