semble
Verifiedby MinishLab
Search a codebase the way an agent needs to — semantic and structural queries that return the relevant spans directly, cutting token use sharply versus grep-then-read loops.
- Access
- CLIMCP
- Auth
- None
- Skills
- 3
Skills
Semantic Code Search
Finds code by meaning rather than exact string match, returning ranked spans instead of whole files.
Local Indexing
Builds and refreshes an on-disk index of a repository so searches run locally without sending code to a service.
Token Efficient Output
Returns trimmed, relevance-scoped results designed to keep agent context small on large repositories.
Related Agents
jevgrep
by David Zhang
Finds the files and source excerpts relevant to a plain-language question about a repository, giving coding agents a…
Codebase Memory MCP
by DeusData
Indexes repositories into a persistent tree-sitter knowledge graph of functions, classes, call chains, and HTTP routes…
Serena
by Oraios AI
IDE-like MCP toolkit for coding agents — semantic code retrieval, editing, refactoring, and debugging with language…
Maki
by Maki
Rust coding agent CLI with tree-sitter code indexing, sandboxed Python execution, and multi-tier subagents. Benchmarks…