masihmoloodian

masihmoloodian / sema

Public

Semantic codebase indexer and MCP server for Claude Code.

11
0
85% credibility
Found May 19, 2026 at 12 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Sema is a local code search tool that makes AI coding assistants like Claude Code work faster on large projects. Instead of making the AI read through entire files to find things, Sema indexes your codebase once and gives the AI a fast search interface. Everything runs on your own computer—no code ever leaves your machine. The main benefit is saving time and AI usage costs: questions that used to cost thousands of tokens now cost hundreds. Sema supports TypeScript, Python, Go, and many other languages, and integrates with Claude Code through a simple setup process.

How It Works

1
💭 You have a big project and Claude is slow

Every time you start a new chat, Claude spends minutes reading files just to understand your project before helping.

2
🔍 You index your entire codebase once

Sema scans every function, class, and method in your project and builds a searchable map stored right on your computer.

3
🔌 You connect it to Claude Code

A simple setup command registers Sema as a helper that Claude can call whenever it needs to find something in your code.

4
You ask Claude a question
🎯
Backend questions

Claude searches for relevant functions and classes, then shows you exactly what you need

🗺️
Architecture questions

Claude shows you a map of your project structure and all exported symbols

5
⚡ You get answers instantly

Same answers as before, but 10 times faster and using far fewer tokens. No more watching Claude fumble through files.

🎉 Your coding sessions are smoother

Claude starts helpful from the first message. You save time, save money on AI usage, and finally have a smart assistant that knows your codebase.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 12 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is sema?

Sema is a local semantic code indexer that gives Claude Code a searchable map of your entire codebase. Instead of Claude burning thousands of tokens reading files to understand your project from scratch each session, sema pre-indexes every function, class, and method so Claude can search directly for what it needs. It runs entirely on your machine using tree-sitter for parsing, SBERT embeddings for semantic search, and ChromaDB for storage. You index once with `sema index .`, register it as an MCP server with `sema init`, and Claude gains five new tools: `search_code`, `get_code`, `repo_map`, `find_usages`, and `explain_file`.

Why is it gaining traction?

The hook is simple: token savings. The README shows a concrete example where a single "how does auth work?" question costs 9,000 tokens without sema and under 900 with it. That's a 10x reduction in AI costs for the navigation phase alone. Developers working in large TypeScript or Python monorepos feel this pain immediately at the start of every Claude Code session. The fact that it requires no API keys, no external services, and no GPU makes it trivial to try. The MCP integration means it slots into Claude Code without any custom prompting or workflow changes.

Who should use this?

Backend and full-stack developers working in TypeScript, Python, or Go monorepos who use Claude Code daily will see the most benefit. If your project spans 50+ files and you find yourself waiting for Claude to explore before it helps, sema addresses that directly. Teams using Claude Code for code review, refactoring, or architectural questions will recover the most tokens since those tasks require broad codebase navigation. Early-stage projects with simple structures will see less value since the index-once overhead may not pay off.

Verdict

At 11 stars with an experimental disclaimer, sema is a promising prototype with a sharp problem definition and solid technical choices. The credibility score of 0.85% reflects the project's youth, but the documentation is thorough, the CLI is well-designed, and the MCP tool interface is clean. Worth installing to test on your largest codebase today if you use Claude Code regularly -- just don't rely on it for production workflows until incremental indexing lands in v0.3.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.