All releases
Feature

Auth: let AI agents act for your users

Your users can connect Claude Code or another AI agent to your app with one command and approve it in their browser. Auth runs the OAuth 2.1 server; the agent gets a short-lived token for your MCP server only, acts within the user's role, and stops the moment they disconnect it.

Turn on agent access for an Auth app and point it at your MCP server and your consent page. When a user runs claude mcp add with your server's URL, their browser opens your app, they sign in as usual and choose what the agent may do. Auth handles client registration, PKCE, tokens bound to your server, refresh tokens that rotate on every use, and revocation. Each user sees their connected agents and can disconnect one at any time; every grant and revocation is in your app's audit log.

dispute.markets is the first app to use it. Read Agent access to set it up.