GH Repository · MinishLab
semble
Fast and Accurate Code Search for Agents. Uses 99% fewer tokens than grep+read
- stars
- 6,100
- 30-day movement
- +3411/day
- Related entries
- 60
- Connections
- 1
python/uvpythonmakePythonretrievalmodel-context-protocolmcp-servermcpembeddingscode-searchagents
semble is an MCP server from Minish Lab that provides fast, accurate code search for agents. It uses embeddings-based retrieval and claims to use 99% fewer tokens than a grep-plus-read workflow.
Reach for it when an agent needs to find relevant code in a repository without burning tokens on broad grep output and full file reads.
Use it to
- Wire code search into an agent via MCP
- Replace grep+read loops with semantic retrieval
- Locate relevant code snippets in large codebases
- Reduce token usage in coding agent workflows
For Developers building coding agents that search codebases
- Role
- mcp-server
- Language
- Python
- Licence
- MIT
- Forks
- 268
- Open issues
- 3
- Last push
- 2026-09-12
- Latest release
- v0.1.0 · 2026-04-26
topicscode-searchembeddingsmcp-serveragentsretrievalpython