artificery/dew
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
dew/docs/features/mcp.md
Chris Hendrickson 0723c7d996 Scaffold docs, packages, and command registration 
- Flesh out docs/index.md, docs/config.md, docs/features/mcp.md,
  docs/features/kanban.md with full content and aligned tables
- Fix .markdownlint-cli2.mjs config (rules were outside 'config' key)
- Add packages/mcp as a standalone MCP server package
- Update all pubspec descriptions
- Implement DewCommand + CommandRegistry in core
- Implement KanbanCommand and McpCommand stubs with registerCommands()
- Wire CLI entry point using CommandRunner + CommandRegistry
- Add 'melos run dew' script for running the CLI from workspace root
- Update tests across core, kanban, mcp packages

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-04-23 14:29:09 -04:00

1.1 KiB
Raw Blame History

Dew Model Context Protocol (MCP) Server

The Dew Model Context Protocol (MCP) Server is a feature that allows AI agents to interact with your project. This enables you to integrate AI capabilities into your project management workflow, such as automated task creation, progress tracking, and more.

Package structure

The MCP feature is split across two packages to keep concerns separate:

  • packages/core defines the McpToolProvider interface. Any feature package that wants to expose tools to AI agents implements this interface — without needing to depend on the MCP server itself.
  • packages/mcp implements the actual server. It collects all registered McpToolProvider implementations and serves them over the configured host and port. Only the cli package depends on packages/mcp; feature packages like kanban remain decoupled from the transport layer.

Configuration

The MCP server is configured under the mcp key in .project/dew.yaml. By default it runs on localhost at port 8080.

dew:
  mcp:
    host: "localhost"
    port: 8080

See the Configuration documentation for full details.