kimi-code-mcp
Provides an MCP server that lets Claude Code delegate bulk codebase analysis to Kimi Code (Kimi K2.5, up to 256K context) by spawning the Kimi CLI and returning parsed JSON results via MCP tools such as kimi_analyze and kimi_resume.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
TLS likely used for provider API calls (api.kimi.com over HTTPS). Secrets are recommended via environment variables and config references rather than hardcoding; however, the README does not describe how the MCP server handles/logs subprocess output, redacts secrets, or implements request/response filtering. No scope/granularity details are provided at the MCP server layer.
⚡ Reliability
Best When
You have large, multi-file repositories and want Claude to reason/edit using compressed, cross-file analysis output rather than reading everything itself.
Avoid When
When you need every line verbatim from the analysis (detailed output approaches raw reading costs) or when external execution/subprocesses are disallowed.
Use Cases
- • Delegate large codebase architecture and dependency analysis to Kimi, then have Claude act on the report
- • Security/vulnerability review over an entire repository using structured findings from Kimi
- • PR/pre-merge review workflows where a bulk scan precedes targeted edits
- • Session-based analysis with resume and cached context to reduce repeated reading
- • Lightweight Q&A against a repository without full bulk context (kimi_query)
Not For
- • Small repos or single-file changes where direct reading by Claude is faster
- • Use cases requiring a first-party REST/GraphQL API surface for direct client integration
- • Environments that cannot run subprocesses (the MCP server shells out to the Kimi CLI)
Interface
Authentication
Authentication to this MCP server is indirect: it relies on local Kimi CLI authentication/config (OAuth or API key). Scopes are not described at the MCP server layer.
Pricing
Cost savings described as reduced Claude token consumption; actual spend depends on Kimi plan and usage frequency.
Agent Metadata
Known Gotchas
- ⚠ Requires valid Kimi Code membership and a completed `kimi login` (or equivalent API key configuration) because the MCP server invokes the Kimi CLI.
- ⚠ The MCP tool name is `kimi_analyze`, but the underlying CLI uses flags like `--work-dir`, `--print`, `-p/--prompt` (no `kimi analyze` subcommand).
- ⚠ Non-interactive execution depends on subprocess behavior; failures may be surfaced via MCP without clear documented mapping.
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for kimi-code-mcp.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.