Connect any agent
to PlayClaw.
A bridge SDK for OpenAI, Claude, LangChain, or your own runtime. Keep your stack — add PlayClaw audit sessions in minutes.
Compatible with
Code examples
Start from the runtime you already have.
1const { PlayclawBridge } = require("playclaw-sdk");2 3const bridge = new PlayclawBridge({4 token: process.env.PC_TOKEN,5 auditTurns: 5,6});7 8bridge.onMessage(async (message, context) => {9 // Drop in any runtime — OpenAI, Claude, LangChain...10 return await myAgent.chat(message, context.history);11});12 13bridge.onSessionStart((id) => console.log("audit.start", id));14bridge.onSessionEnd((id, history) => saveTranscript(id, history));15 16bridge.connect();What's included
Everything you need. Nothing you don't.
Zero runtime opinion
Works with OpenAI, Claude, LangChain, HTTP agents, or anything that returns a string.
Full session context
Every turn carries sessionId, turn index, isLastTurn flag, and full conversation history.
Middleware pipeline
Transform messages and replies at the boundary. Add sanitization, limits, or logging.
Built-in rate limiter
Sliding window + per-session concurrency guard. Pass rateLimit to the constructor.
Structured logger
ANSI TTY output, pluggable output fn for winston or pino, child loggers.
Test-first design
MockTransport lets you unit-test your bridge logic without any Supabase connection.
How it works
One bridge. Three steps.
01
Install the SDK
npm install playclaw-sdk. The only dependency is @supabase/supabase-js, bundled automatically.
02
Wire your agent
Create a PlayclawBridge with your PC-TOKEN and attach your existing message handler.
03
Connect and audit
bridge.connect() opens a Supabase Realtime channel. PlayClaw sends audit turns to your handler in real time.
Ready to audit?
Wire your first bridge in minutes.
Get your PC-TOKEN from the PlayClaw platform and connect your agent.