code-assistant and codebase-mcp
These are complements: stippi/code-assistant provides the autonomous agent framework while danyQe/codebase-mcp supplies the MCP protocol implementation for grounding that agent in local codebase context via semantic search.
About code-assistant
stippi/code-assistant
An LLM-powered, autonomous coding assistant. Also offers an MCP and ACP mode.
Built in Rust, it supports pluggable tool invocation modes (native function calling, XML tags, triple-caret blocks) for compatibility across AI providers, while streaming tool invocations in real-time with smart safety filtering. Provides GUI (via Zed's GPUI), CLI, and headless MCP/ACP modes for direct integration with Claude Desktop and ACP-compatible editors like Zed. Sessions persist state with automatic context compaction, format-on-save capabilities, and auto-loads repository guidance from `AGENTS.md` files to align assistant behavior with project-specific instructions.
About codebase-mcp
danyQe/codebase-mcp
Open-source AI development assistant via Model Context Protocol (MCP). Turn Claude or any LLM into your personal coding assistant. Privacy-first with local semantic search, AI-assisted editing, persistent memory, and quality-checked code generation. Built for Python & React. Free alternative to paid AI coding tools.
Based on the README, here's a technical summary that goes beyond the GitHub description: Implements a FastAPI backend server that exposes 13+ MCP tools via stdio transport, with a local FAISS vector index for sub-second semantic code search and persistent memory storage via SQLite. Leverages Google's free Gemini API exclusively for AI-assisted code edits while keeping all other operations local—semantic search, memory recall, and file writing happen entirely on-device.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work