Aira exposes its product management capabilities through the Model Context Protocol (MCP). This lets Claude Code, Claude Desktop, ChatGPT, and any other MCP-compatible client call Aira as a tool server.
What MCP gives you
Instead of switching to the Aira web UI, you can do PM work directly from your AI assistant:
- Create and manage projects
- Add sources and trigger analysis
- Generate insights, PRDs, features, and sprint plans
- Assign and update tasks
- Chat with Aira for status queries and reports
- Monitor project health via pulse actions
The web UI and MCP use the same backend logic. Behavior is consistent across both interfaces.
Three deployment modes
| Mode | URL | Auth | Best for |
|---|---|---|---|
| Managed Remote | mcp.aira.pro | OAuth (browser login) | Most users — no setup required |
| Self-hosted Remote | mcp.your-domain.com | OAuth (your instance) | Enterprise, private data, compliance |
| Local | localhost:8000 | API key | Development, debugging, no internet |
When to use each
- Managed Remote — You have an Aira account at
app.aira.proand want to connect Claude or ChatGPT. Start here. - Self-hosted Remote — You're deploying Aira in your own cloud for security or compliance reasons. Your MCP clients connect to your instance.
- Local — You're developing against aira-agent locally and want to test MCP integration. Everything runs on your machine.
Structured-first usage
When calling Aira via MCP, prefer typed tools over the chat fallback:
- Use
update_task,assign_task,start_sprint,generate_pulse_actionsfor state changes - Use
list_tasks,list_features,get_current_sprintfor queries - Fall back to
chat_with_airaonly when a typed tool doesn't exist for your intent
Guides
- Managed Remote — Connect to
mcp.aira.pro(recommended) - Self-hosted Remote — Deploy your own MCP server
- Local — Run everything on your machine