MinishLab/semble is a Python-based code search library designed to provide fast and accurate code snippet retrieval for agents, significantly reducing the token usage compared to traditional methods like grep+read.
Source: Official README and linked documentation View on GitHub →This project is being discussed because its significant reduction in token usage for code search, its fast indexing and querying capabilities, and its compatibility with various coding agents and MCP servers. The unique technical choice of using embeddings and a specialized search algorithm differs, offering a more efficient and user-friendly code search experience.
Source: Official README and linked documentationSemble indexes and searches codebases quickly, with an average indexing time of ~500 ms and query response time of ~1 ms, while maintaining high accuracy in code snippet retrieval.
Source: Official README and linked documentationSemble returns only relevant code snippets, using approximately 99% fewer tokens than grep+read, making it more efficient in terms of resource usage.
Source: Official README and linked documentationSemble operates on CPU without requiring API keys, GPUs, or external services, making it easy to set up and use.
Source: Official README and linked documentationSemble can be used as an MCP server for direct agent integration and as a CLI tool for standalone searches, offering flexibility in usage scenarios.
Source: Official README and linked documentationThe architecture of Semble is inferred to be modular, with clear separation between indexing, searching, and user interface components. It likely employs design patterns such as the Model-View-Controller (MVC) for managing data flow and user interactions. The use of embeddings and a specialized search algorithm suggests a focus on efficient and accurate code snippet retrieval.
Source: Official repository tree and dependency filesmodel2vecvicinitynumpypathspecorjsonquestionarysemble-grammarsSemble is suitable for developers and technical teams who need efficient code search capabilities within their development environments, particularly for use with coding agents and MCP servers. It is useful in scenarios where quick access to relevant code snippets is critical, such as during code reviews, debugging, or when integrating new code into existing projects.
Source: Official README and linked documentationThe latest GitHub release is v0.6.0. Its release page records multi-repository search, CLI progress and formatted output, and a fix for chunk line numbers. The v0.2.0 value in older drafts is outdated.
Source: github.com/MinishLab/semble/releases/tag/v0.6.0；https://github.com/Mi…status: docs-only | checked_at: 2026-09-23 | environment: Docs only. Reviewed the official MinishLab/semble README, pyproject.toml, installation documentation, and Releases. | limitations: uv and semble were not installed, no model was downloaded, no repository was indexed, no benchmark was run, and official performance figures were not verified locally. | observed_output: pyproject requires Python 3.10 or newer. The README currently recommends uv tool install semble followed by semble install. GitHub Releases marks v0.6.0 as Latest.
Source: github.com/MinishLab/semble；https://github.com/MinishLab/semble/blob/…ripgrep performs line-based regular-expression search and follows gitignore by default, making it suitable for exact and exhaustive text matching. Semble maps natural-language intent to ranked code snippets and helps narrow the search when the symbol name is unknown. A practical investigation can use Semble for direction, followed by ripgrep to verify every literal match.
Source: github.com/BurntSushi/ripgrep；https://github.com/MinishLab/sembleMinishLab/semble is a project for any team or individual seeking an efficient and accurate code search solution. Its focus on reducing token usage and providing fast access to code snippets makes it particularly valuable for developers working with large codebases or integrating with coding agents.
pyproject requires Python 3.10 or newer. The README currently recommends uv tool install semble followed by semble install. GitHub Releases marks v0.6.0 as Latest.
The review environment was: Docs only. Reviewed the official MinishLab/semble README, pyproject.toml, installation documentation, and Releases. semble behavior outside that scope still needs a local check.