MoonshotAI / MoonshotAI/kimi-code
[Feature Request]Support MCP tools for Plan Mode
Nobody has claimed this yet.
- Dominant language
- TypeScript
- Stars
- 7.5k
- Forks
- 1.2k
- Avg merge
- 11h 53m
- Merged PRs (30d)
- 350
Description
What feature would you like to see?
Background
There are a large number of existing MCP servers, like codegraph and codebase-memory-mcp, designed to cut down token consumption when agents read and analyze source code. These MCP tools greatly optimize code context loading, avoid sending massive raw code snippets into prompts, and effectively reduce token usage.
Current Limitation
The current Plan Agent cannot connect or leverage any MCP servers. As a result, we are unable to utilize those token-saving MCP tools during code planning and code analysis workflows.
https://github.com/MoonshotAI/kimi-code/blob/44f334191989183d21920f6867c405581347c748/packages/agent-core-v2/src/agent/plan/profile/plan.ts#L20-L27
Requested Feature
Add native support for MCP to the Plan Agent module, so Plan Agent can connect to, call, and consume existing code-optimized MCP servers for code reading and analysis tasks.
Use Case
When the Plan Agent needs to parse project structures, read file contents, or inspect codebases during task planning, it can offload code fetching and preprocessing to lightweight MCP tools to lower overall token overhead.
Additional information
No response
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with packages/agent-core-v2/src/agent/plan/profile/plan.ts, especially lines 20-27, and trace the Plan Agent entry point. Review how MCP servers are connected and invoked elsewhere in the repository before defining the integration. Done means the Plan Agent can connect to, call, and consume existing MCP tools during code planning and analysis.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- typescript
- Domain
- ai, cli
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100