Search
Give your coding agent fast, structured ripgrep and fd searches so it spends fewer tokens finding symbols and files in your repo.
Overview
io.github.Dave-London/search is a MCP server for the Build phase that runs ripgrep and fd code search and returns structured, token-efficient results to your agent.
What is this MCP server?
- Combines ripgrep and fd for pattern and path discovery with MCP tool calls
- Returns structured, token-efficient results instead of dumping raw terminal walls
- Ships as @paretools/search over stdio (npm registry v0.7.1)
- Solo builders use it to shrink context when agents grep large monorepos
- Part of the Pare toolbox alongside other Dave-London MCP utilities
- Published npm package @paretools/search at version 0.7.1
- Uses stdio MCP transport per server schema
- Backed by ripgrep and fd as stated in the server description
Community signal: 128 GitHub stars.
What problem does it solve?
Agents waste context and time when code search returns noisy ripgrep dumps or when you cannot expose fast fd path queries through MCP.
Who is it for?
Solo builders with medium or large repos who want MCP-native search that stays cheap on tokens during feature work and refactors.
Skip if: Teams that only need occasional human IDE search and do not run agents against the repo, or anyone expecting semantic/embeddings search without ripgrep or fd.
What do I get? / Deliverables
After you register @paretools/search, your agent can query the codebase with compact structured hits instead of hand-rolling grep in every session.
- Structured code and path search results via MCP tools
- Lower-token search transcripts for agent sessions
- Repeatable repo exploration without manual shell paste
Recommended MCP Servers
Journey fit
Code navigation and discovery happen constantly while you implement features, wire integrations, and refactor—this server belongs on the Build shelf as agent tooling. It is an MCP bridge for search primitives (ripgrep + fd), not a ship-time test runner or an operate monitor—canonical placement is agent-tooling under Build.
How it compares
MCP code-search integration built on ripgrep and fd, not a semantic indexing service or a documentation-only skill.
Common Questions / FAQ
Who is io.github.Dave-London/search for?
It is for solo and indie developers using Claude Code, Cursor, or Codex who want token-efficient ripgrep and fd search exposed as MCP tools.
When should I use io.github.Dave-London/search?
Use it during Build when an agent needs to locate symbols, strings, or files repeatedly without pasting large grep output into the chat.
How do I add io.github.Dave-London/search to my agent?
Install @paretools/search from npm, configure the MCP server with stdio transport in your agent’s MCP settings, and point it at your project root like any local dev tool.