semble
Fast semantic code search for agents: returns the exact snippets instead of reading whole files and spends far fewer tokens
Medium risk
We rate an entry medium when the tool runs code, makes network calls or reads project files. Check what exactly it does before installing.
Why this level
- Reads and indexes repository files locally
- Runs a local service for search
Install
Manual install
uv tool install semble
semble installThe installer detects agents and configures their connection.
This is third-party code. Review the repository files before installing.
What it does
Semble indexes a repository and, given a natural-language query, returns the exact code snippets rather than whole files. It runs on CPU with no keys, GPU or external services, and indexes and searches most projects in seconds. An agent connects Semble as an MCP server, as a CLI through AGENTS.md, or as a dedicated sub-agent. An interactive installer detects installed agents such as Claude Code, Cursor and Codex and configures them.
Who it is for. For developers who let an agent quickly navigate a large or unfamiliar codebase.
Good fit when
- The agent greps and reads whole files, burning context
- You need fast semantic search over a large repository
- You need to find where a feature or handling lives by description
Not a fit when
- The project is tiny and plain search is enough
- You need search over documents or a database, not code
- You cannot install local tools via uv
Example request
Find where user authentication is handled in this repositoryLimitations
Requires uv to install. The index is built locally and needs disk and memory scaled to the repository size. Search quality depends on the language and code structure.
How to disable. Remove semble from the MCP config or AGENTS.md and uninstall the tool with uv tool uninstall semble.
MCP
- Transport
- stdio
- Authentication
- not required
Security check
- Reads and indexes repository files locally
- Runs a local service for search
README in short
Semble is a semantic code-search library and MCP server for agents from the MinishLab team. It returns relevant snippets, saving context compared with grep and file reads, and indexes repositories much faster than code-specialized transformers. Everything runs on CPU with no keys or external services. It installs via uv, and semble install wires up detected agents. MIT licensed.
FAQ
Do I need keys or a GPU?
No, Semble runs on CPU locally with no keys or external services.
Which agents work with it?
Any coding agent that supports MCP, a CLI via AGENTS.md, or sub-agents, including Claude Code, Cursor, Codex and OpenCode.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything