The loop · 01 Connect · No webhook? Start here

Stand up a conforming agent in a minute.

No endpoint yet? Pick a managed-LLM connector template and download a ready-to-run server that already passes the ship-gate contract — conformant by construction. Host it anywhere, set your own credentials, and connect it back. You never write a webhook.

Probe-conformant by constructionSecret-free — your keys stay on your hostNode 18+ · one file to edit

Name your agent (optional)We bake it into the starter as my-agent. You can rename it after download.

Managed

OpenAI Assistant proxy

Proxy an existing OpenAI Assistant — threads, runs, and polling handled.

You bring: An OpenAI API key + Assistant ID
Env: OPENAI_API_KEY · OPENAI_ASSISTANT_ID · DISPATCH_TOKEN · PORT
Deps: express · openai

Managed

Anthropic agent proxy

A Claude agent loop with a tool registry you extend — tool-use handled.

You bring: An Anthropic API key
Env: ANTHROPIC_API_KEY · ANTHROPIC_MODEL · DISPATCH_TOKEN · PORT
Deps: express · @anthropic-ai/sdk

Managed

LangGraph / CrewAI proxy

Forward each run to your LangGraph or CrewAI deployment and map the result back.

You bring: Your deployment URL (+ optional bearer token)
Env: AGENT_BACKEND_URL · AGENT_BACKEND_TOKEN · DISPATCH_TOKEN · PORT
Deps: express

Direct

Anthropic (Claude) — direct

Wrap Claude directly on your own key. The default scaffold.

You bring: An Anthropic API key
Env: ANTHROPIC_API_KEY · LLM_MODEL · DISPATCH_TOKEN · PORT
Deps: express · @anthropic-ai/sdk

Direct

OpenAI (GPT) — direct

Wrap a GPT chat model directly on your own key.

You bring: An OpenAI API key
Env: OPENAI_API_KEY · LLM_MODEL · DISPATCH_TOKEN · PORT
Deps: express · openai

Already hosting an endpoint? Connect it directly →