Agent-to-Agent

Let your AI tools talk to Chalie — ask questions, share context, and coordinate work.

Your AI tools can talk to Chalie

Chalie exposes an MCP server that lets other AI agents communicate with it directly. If you use Claude Code, Cursor, Windsurf, or any MCP-compatible tool, it can ask Chalie questions, share updates, and request actions — just like you do in conversation.

Why this matters

Your coding assistant knows your codebase. Chalie knows your calendar, email, contacts, documents, and personal context. Agent-to-agent communication bridges the gap:

  • Claude Code can ask Chalie about your schedule before suggesting a meeting time
  • A CI bot can tell Chalie about a failed deployment so it can notify you with context
  • Any MCP client can access Chalie’s full set of capabilities — memory, email, calendar, lists, and more

How it works

Chalie runs an MCP server alongside its main interface. The server uses Streamable HTTP transport with bearer token authentication. External agents connect over your local network and communicate through a single tool called talk_to_chalie.

When an agent sends a message, Chalie processes it with the same intelligence it uses for your conversations — it has access to all your memories, documents, and connected services. The agent gets back a direct response.

Privacy and control

  • The MCP server only runs on your local network — nothing is exposed to the internet
  • Every connection requires a bearer token you generate from the Brain dashboard
  • You control whether exchanges are visible to you via the loop_in_human flag
  • When loop_in_human is true, Chalie shows you what was discussed

Getting connected

Head to the Setup guide for step-by-step instructions on connecting your tools, then see Usage for practical examples.