fff
The fastest and the most accurate file search SDK for AI agents, Neovim, Rust, C, Python, Bun and NodeJS
FFF is an MCP server that provides high-performance, frecency-ranked file search capabilities to AI agents and coding assistants. It offers typo-resistant fuzzy matching for both file paths and content, maintaining a long-lived in-memory index that delivers sub-10ms search results compared to multi-second ripgrep invocations. The server exposes ffgrep, fffind, and fff-multi-grep tools that work with Claude Code, Codex, Cursor, Cline, and any MCP-compatible client, reducing token waste and improving agent response times through smart-case matching, definition-first hinting, and git-aware annotations.
Key Features
Use Cases
- 01Accelerate AI agent file searches in large codebases (500k+ files) from seconds to milliseconds
- 02Enable context-aware code navigation in Claude Desktop or Cursor with frecency-ranked results
- 03Reduce wasted LLM context by surfacing git-modified and definition-containing files first
- 04Support multi-pattern searches in code review or refactoring tasks without spawning multiple processes
- 05Provide typo-tolerant content grep for exploratory agent queries with fuzzy fallback
- 06Integrate real-time file search into long-running agent harnesses or IDE extensions
Related MCP Servers
View moremarkitdown
Python tool for converting files and office documents to Markdown.
firecrawl
The context API to search, scrape, and interact with the web at scale. 🔥
prompts.chat
f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.
langflow
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
fff — FAQ
What is the FFF MCP server?+
The FFF MCP server exposes fast, frecency-ranked file and content search tools (ffgrep, fffind, fff-multi-grep) to AI agents via the Model Context Protocol. It maintains a long-lived in-memory index that delivers sub-10ms searches on large repositories, replacing slow per-query ripgrep spawns.
How do I install the FFF MCP server?+
Run the one-line installer for your platform: curl -L https://dmtrkovalenko.dev/install-fff-mcp.sh | bash on Linux/macOS, or irm https://raw.githubusercontent.com/dmtrKovalenko/fff/main/install-mcp.ps1 | iex on Windows PowerShell. Alternatively, install via Homebrew with brew install dmtrKovalenko/fff/fff-mcp. The scripts print wiring instructions for your MCP client.
Which AI clients and editors support FFF?+
FFF works with Claude Code, Codex, OpenCode, Cursor, Cline, Claude Desktop, and any MCP-capable client. Special setup instructions are provided for Codex, which requires an absolute path in ~/.codex/config.toml.
Does FFF require API keys or external services?+
No, FFF is a local-only tool with no external dependencies or API keys. It indexes your repository on disk and keeps the search state in memory.
Is FFF free and open source?+
Yes, FFF is MIT-licensed and open source forever. All binaries and packages are freely available on GitHub releases and npm.
Why use FFF instead of ripgrep or fzf for AI agents?+
Ripgrep and fzf spawn new processes for each search, re-reading .gitignore and rebuilding state every time. FFF keeps a warm in-memory index, reducing 3-9 second ripgrep invocations to sub-10ms queries on large repos, and returns typed objects with git status and frecency scores instead of unparsed text.
How do I install fff?+
Open the source repository on GitHub and follow its README. fff is a mcp server — MCP Agents Market links you directly to the official repo.
Is fff free?+
fff is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.