FI-Mihej/codebase-agent-mcp

Overview

A sub-harness that acts as both an MCP server and client. It delegates documentation and source code analysis, as well as interactions with related context-providing MCP servers, to a local or inexpensive OpenAI-compatible LLM. This reduces token usage and context size for coding agents running on a top-tier LLM.

Best for

Best for
Developers using top-tier coding agents who want to cut token costs by offloading analysis to cheaper models.

Use cases

• Offload codebase analysis to a cheaper local LLM
• Provide context from MCP servers without consuming top-tier tokens
• Reduce overall token costs for AI-assisted development

Notes

A sub-harness that acts as both an MCP server and client. It delegates documentation and source code analysis, as well as interactions with related context-providing MCP servers, to a local or inexpensive OpenAI-compatible LLM. This reduces token usage and context size for coding agents running on a top-tier LLM.

1 stars on GitHub. Last updated 2026-08-02. Licensed Apache-2.0.

Use cases

• Offload codebase analysis to a cheaper local LLM
• Provide context from MCP servers without consuming top-tier tokens
• Reduce overall token costs for AI-assisted development

Pros

• Lowers token consumption for expensive coding agents
• Leverages local or low-cost LLMs for analysis tasks
• Integrates as both server and client in MCP workflows

Cons

• Limited to analysis and context delegation, not full coding
• Depends on an external LLM for the delegated work
• May add latency or complexity in setup

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.