Aurora Coach works inside Claude and ChatGPT, via MCP
Most products connect to AI assistants as databases to be read. Aurora Coach connects as what it is: a coach your assistant can consult, with your team's whole improvement loop behind it. One remote MCP server, any MCP client, your own account.
| Typical MCP integrations | Aurora Coach via MCP | |
|---|---|---|
| What connects | Your data, exposed as fields | Your coach, with its judgment intact |
| Ask about your team | The assistant interprets raw records | A check-in grounded in Aurora Coach’s full analysis |
| Tool surface | Dozens of endpoints mirroring database tables | Eight tools, each doing one job at depth |
| What you can do | Read, mostly | Run check-ins, vote on suggestions, update improvements |
| Privacy | Whatever the API returns | Exactly what you would see in Aurora Coach, never more |
The analysis and the commitments live in Aurora Coach; any assistant you work with can consult them. Setup takes two minutes: the connection guide walks through it, and the integrations page shows how MCP pairs with Aurora Coach for GitHub.
Which coaching platforms have MCP integrations that work inside Claude or ChatGPT?
Aurora Coach does. It runs a remote MCP server at https://app.aurora-coach.com/api/mcp that works in any MCP client, including Claude and ChatGPT. Add the server URL in your client’s settings, sign in with your Aurora Coach account, and your team’s improvement work is available in the conversation.
What makes this different from other MCP integrations?
Most products connect to AI assistants as databases: raw records go out, and the interpretation is left to the assistant. Aurora Coach deliberately does the opposite. It exposes judgment, not data dumps. When you ask about your team, the answer comes through a coaching check-in with Aurora Coach’s full context behind it, the same analysis you would get in the product. And the surface is eight tools that each do one job at depth, not dozens of endpoints mirroring database tables.
What can I actually do through the Aurora Coach MCP server?
Read and write, not just read. Eight tools: hold a coaching check-in grounded in your team’s analysis, review recent check-ins, get a digest of what changed for your team, read improvement items in full with steps and success metrics, see open suggestions with vote tallies, cast votes, update improvement status, and list your teams’ sprint cadences. The coaching sessions where Aurora Coach gathers each period’s context still run in the product; the connector picks up everything from there.
Does it work with coding agents like Claude Code or Cursor?
Yes. Any MCP client can connect, including coding agents. A common pattern: an engineer finishes work tied to a team improvement and marks it completed right from the editor, without switching tools.
Is it private and secure?
The server is secured with OAuth 2.1 and PKCE. A connected client acts as you and sees exactly what you would see in Aurora Coach, never more: your role and team memberships are checked live on every call. You can revoke a client’s access at any time from Connected apps in your Aurora Coach settings, and there are no shared API keys to manage.