fff
A fast file search toolkit for people and AI agents, usable from Neovim, code, or an MCP server.
- Stars
- 11.2k
- Forks
- 477
- Licence
- MIT
- Language
- Rust
- Last update
- 5 days ago
- Type
- Library
What it is
FFF is a file search toolkit that finds files by path and searches file contents, tolerating typos and ranking files by how often you use them. It keeps a lightweight in-memory index with a background watcher, so repeated searches in a long-running process are fast. It ships as an SDK for several languages, a Neovim plugin, and an MCP server that AI coding agents can call.
- Give an AI coding agent fast path and content search over a git-indexed project through the MCP server
- Add typo-tolerant file search to a Neovim setup
- Embed file and content search in a Python or Node.js tool
- Search for declarations in a codebase with the definition search option
Great for
- Developers building an AI coding agent or editor that needs fast file search
- Neovim users who want fuzzy file and content search
- People running an MCP-capable client who want agents to search a codebase through a resident index
- Teams embedding file search in Rust, Python, Node.js, or Bun applications
Not ideal for
- Someone who only needs a one-off search in a terminal, since the README says the gains come from long-running processes that search more than once and plain ripgrep or fzf may be enough
- Users who want a polished end-user desktop app, because the README describes a toolkit and SDK rather than a standalone app
How to get it
- 1
Install with a package manager
You'll need: Homebrew (for the Homebrew install route)
Homebrew install of the MCP server
brew install dmtrKovalenko/fff/fff-mcpRegister the Homebrew-installed server with Codex
codex mcp add fff -- "$(brew --prefix)/bin/fff-mcp" - 2
Run it
- Install the fff-mcp binary using the one-line installer or Homebrew.
- Register the server with your MCP client, using the absolute path to the binary.
- Restart the client or start a new task so it loads the server.
- Ask the agent to "use fff" and confirm it calls the grep, find_files, or multi_grep tools.
- 3
Get the most out of it
- Pass a known directory or module in the first search using the path option to narrow scope.
- Use definitions_only to search declarations rather than all matches.
- Use mode regex or literal with paths and globs for ripgrep-style searches.
- Add the recommended prompt from the README to your CLAUDE.md or equivalent file.
Downloads
Strengths
- The README describes benchmark results for MCP versus shell search, with the setup and test conditions stated
- Available across several languages and editors, with MIT license
- Actively maintained, with a recent push and a published release
Trade-offs
- Benchmark figures in the README come from the project's own tests and their specific repositories and models, so real gains will vary
- Many options (globs, modes, context, stages) make the full behavior complex to learn from the README alone
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| fff (this) | 11.2k | Library | intermediate | 5 days ago | Developers building an AI coding agent or editor that needs fast file search |
| fzf | 83.9k | Dev tool | beginner | 8 days ago | Command-line users who often search file names or command history |
| telescope.nvim | 19.8k | Library | intermediate | 10 days ago | Neovim users who want fast fuzzy search for files and text |
| fzf.vim | 10.3k | Library | beginner | 1 months ago | Vim or Neovim users who already use fzf and want ready-made search commands |
| oh-my-pi | 37.5k | CLI | intermediate | 4 days ago | Developers who want an AI agent working inside a terminal on their own codebase |
| Archon | 23.8k | Dev tool | intermediate | 4 days ago | Developers who want AI coding runs to follow the same plan, test and review steps each time |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.