M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
F

FileScopeMCP

Community
Your AI already knows how to code. Now it knows your codebase.
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #176 of 438
Stars
★ 302 Popular
Transport
stdio (local process)
Runtime
Node.js 22+
Credentials
No credential needed
License
Other / unspecified
Last commit
Tools
17
56FMRS · C

FileScopeMCP is a powerful code analysis tool designed to enhance AI assistants' code understanding. It provides precise context through importance scoring, dependency mapping, and symbol extraction, reducing the need for AI to read source code. Installation is straightforward, with support for multiple languages and clients. Its utility is broad, but it's important to be aware of the heuristic nature of scoring, and the privacy/configuration complexity when using local LLMs.

Strongest · Documentation 14/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
12/20
Maintenance
11/20
Documentation
14/20
Setup experience
11/20
Why each score
Reliability 8/20
The repository includes a GitHub Actions build workflow, indicating CI practices. However, no committed tests covering the server's key paths were found in the source; tool behavior is largely based on README claims, and reliance on external LLM components and native modules may cause degradation. Since this is a static review, actual runtime cannot be verified, so score conservatively.
Security and permissions 12/20
The server features a background broker and daemon, using Unix socket for internal communication, with data stored in local SQLite. No obvious credential hardcoding or malicious behavior. However, tools to set file summaries and importance scores may allow data modification; the destructive operation 'exclude_and_remove' is labeled destructive but lacks explicit confirmation. No red-line risks identified.
Maintenance 11/20
The repository is active with recent commits and a build badge, but the license is 'All Rights Reserved', possibly ambiguous. No security response channel or contribution guidelines are provided. Rating slightly above average.
Documentation 14/20
README is comprehensive, covering installation, quick start, MCP tools, architecture, FAQ, and links to several dedicated docs. However, advanced features like Nexus dashboard and monitoring stack may be complex, and tool parameters/limits are not detailed per tool. Documentation quality is good.
Setup experience 11/20
Detailed installation steps are provided (build.sh), requiring Node.js >=22 and build tools, with multi-platform support. Registers via 'claude mcp add' and allows manual configuration. Native modules and LLM dependencies add deployment complexity. Static review lacks execution evidence, so score capped at 15.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesUses the network

Best for

  • Developers who need deep codebase context for AI assistants.
  • Teams that want to maintain code metadata automatically during development.
  • Projects using multiple languages (supports TS/JS, Python, C/C++, Rust, Go, etc.).

Not for

  • Scenarios requiring only a simple file list without dependency analysis.
  • Environments without Node.js >=22.
  • Fully offline settings without an LLM (though basic features still work).

Required permissions

  • File system read and watch permissions (to scan the codebase).
  • Write permissions (to create `.filescope/data.db` database).
  • Network permissions (optional, to connect to a local LLM endpoint).

Risks and side effects

  • Importance scoring is heuristic and may not align with all project preferences.
  • Regex-based parsing for some languages may be less accurate than AST.
  • LLM summaries depend on external services, potentially raising privacy concerns.
  • Database may consume significant space on large codebases.

Setup

Before you start

Runtime：Node.js 22+

  1. Clone the repository: git clone https://github.com/admica/FileScopeMCP.
  2. Enter the directory and run ./build.sh to install dependencies, compile, and register with Claude Code.
  3. For other clients, refer to docs/mcp-clients.md for manual configuration.
claude_desktop_config.json 
{
  "type": "stdio",
  "command": "node",
  "args": [
    "/path/to/FileScopeMCP/dist/mcp-server.js"
  ],
  "env": {}
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

Generate and validate client config

Check that it works

After running ./build.sh from the README, open any project in Claude Code; find_important_files and status should appear in the tool list. Calling status() and getting broker connection and watcher state back confirms the install.

Troubleshooting

  1. Ensure Node.js >=22 is installed and build tools (e.g., `build-essential`) are available.
  2. Check logs; ensure the watcher is running and the `.filescope/data.db` exists.
  3. If LLM summaries are not generated, verify the local LLM endpoint is configured and reachable.
  4. Refer to `docs/troubleshooting.md` for common issues and fixes.

Things to try

Once connected, you can ask your AI assistant things like:

  • Show me the 5 most important files in this project and why.
  • Who calls the function parseConfig?
  • Are there any circular dependencies in this codebase?
  • What is the summary, exports, and dependencies of src/index.ts?

Tools 17

status read-only
View broker connection, queue depth, LLM progress, and watcher state.
find_important_files read-only
List top files by importance score with dependency counts.
get_file_summary read-only
Get comprehensive metadata about a file, including summary, concepts, change impact, exports, deps, and staleness.
list_files read-only
Show full file tree or flat top-N by importance.
find_symbol read-only
Resolve a symbol name to file and line range; supports prefix match.
find_callers read-only
Find all symbols that call a named symbol (TS/JS call graph).
find_callees read-only
Find all symbols that a named symbol calls (TS/JS call graph).
search read-only
Search file metadata across symbols, summaries, purpose, and paths.
Show 9 more tools
list_changed_since read-only
List files changed since a timestamp or git SHA.
get_communities read-only
Get Louvain-clustered file groups by import coupling.
detect_cycles read-only
Find circular dependency chains.
get_cycles_for_file read-only
Find cycles involving a specific file.
scan_all writes
Queue files for LLM summarization via the broker.
set_base_directory writes
Point at a different project directory.
set_file_summary writes
Manually set or override a file's summary.
set_file_importance writes
Manually set a file's importance score (0-10).
exclude_and_remove destructive
Drop files/patterns from tracking (destructive).

Use cases

Enable AI assistants to quickly understand the structure and key files of a large codebase.
Find call relationships and assess impact before refactoring.
Automatically generate and maintain up-to-date file summaries for AI consumption.
Visualize project dependency graphs and community structures.

Supported clients

Claude Desktop
Cursor AI
Codex

Listed from the project's documentation, not tested by this site.

Overview

FileScopeMCP analyzes your codebase, identifying important files based on dependency relationships. It generates diagrams and importance scores per file, helping AI assistants understand the codebase. Automatically parses popular programming languages such as Python, C, C++, Rust, Zig, Lua. It provides tools for querying file summaries, symbol resolution, call graphs, and more.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision fd179e0a93ee Data synced 2026-10-19 Read the FMRS scoring method