Introducing Chat Agent Relay — Open-Source Chat-to-Agent Middleware

March 2026 · Launch post

The problem

Teams ship chat-to-agent flows as point-to-point integrations: a Slack webhook into custom code, then OpenAI, then the Slack API. For a demo, that is enough. Then the scope grows: Microsoft Teams, audit logs for compliance, a different LLM, retries when delivery fails. Each new requirement tends to mean another bespoke path through the codebase. What started as a thin glue layer becomes a rewrite every time the platform or vendor changes.

Our approach

Make the chat-to-agent layer explicit. Instead of ad hoc calls between transports and models, Chat Agent Relay centers a canonical event model: the same sequence of events for every message, no matter which channel or backend you use. Pluggable adapters sit on both sides — channels normalize into canonical events; backends consume them and return structured outcomes. Governance and audit are part of the pipeline, not an afterthought: policy runs before the agent, and an append-only ledger records what happened.

What Chat Agent Relay provides today

Getting started

From zero to a running server in three commands (fill in .env with your keys as described in the docs):

quickstart
$ git clone https://github.com/ChatAgentRelay/ChatAgentRelay.git && cd ChatAgentRelay $ bun install installed dependencies $ cd packages/server && cp .env.example .env && bun run start

What’s next

We are focused on widening the adapter surface and hardening the core:

Try it

Chat Agent Relay is MIT licensed. Star the repo if it is useful, run through Getting Started, and share the project with your team.

← Back to Blog