Log inSubmit server
File AI logo

File AI

Reads local documents through MCP, creating indexed content, searchable anchors, outlines, and contextual excerpts without modifying source files.

InstallRepository
Full README Compare alternatives Embed widget Claim this listing

Install File AI

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "file-ai": {
      "command": "npx",
      "args": [
        "-y",
        "@flyfish-dev/file-ai"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

We couldn’t confirm it starts correctly

We ran the install command but it didn't respond within our test window — this can mean a slow first-time install rather than a real problem.

npx -y @flyfish-dev/file-ai

No response to initialize.

This is an experimental automated check and can have false negatives — missing environment variables, a slow cold install, etc. It doesn’t necessarily mean something’s wrong. Last checked 8d ago.

Overview

File AI MCP server turns local documents into indexed profiles, blocks, chunks, outlines, search results, and anchored context for AI agents. It uses Flyfish File Viewer core for format recognition and exposes read-only MCP tools over stdio or streamable HTTP. Reach for it when an agent needs grounded access to PDFs, office files, spreadsheets, email, archives, or text-based documents without handling each file format directly.

Use cases

  • Analyze local PDFs and office documents before answering questions
  • Search indexed files and cite matching source anchors
  • Retrieve context surrounding a page, slide, worksheet, or query
  • Inspect supported formats and extractor coverage
  • Build document-aware agents without implementing format parsers

Key features

  • Local document indexing and caching
  • Anchor-preserving reads and searches
  • Context retrieval around queries or anchors
  • Outline, profile, and chunk resources
  • Support for common office, email, archive, and text formats
  • Stdio and streamable HTTP transports

Tools5

What File AI lets your agent do. Expand a tool for its parameters.

Self-reported~94 tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 1d ago

Recent health check succeeded.

5/5 checks healthy over the last 42d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 15h ago via OSV.dev · @flyfish-dev/file-ai (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars3Stargazers on the source repository.
npm downloads153Package downloads in the last 30 days.
Last commit4mo agoMost recent push to the default branch.
Install checkInconclusiveDidn't respond in our test window — often a slow first install.
Tools exposed5Callable tools this server registers over MCP.
Directory activity2 viewsConfig copies, upvotes, and views on AllMCPs.

How File AI works

What File AI MCP server does

File AI MCP server gives agents structured, read-only access to local document content. It can inspect a file, build a reusable index, return document structure, search extracted text, and provide nearby content around a result or anchor. Responses preserve source locations through identifiers such as page, slide, worksheet, row range, nested path, or byte and text positions.

The server is designed for document-grounded workflows where an agent needs to refer back to a specific part of a file. Agents should use returned anchorId values when citing claims derived from document content. Source files are not edited or otherwise mutated.

How it works

doc_analyze parses a local path and caches an index. Subsequent operations can use that index or, where supported, work from a path. doc_read retrieves selected blocks, chunks, or anchors. doc_search searches cached material and returns source anchors, while doc_context finds surrounding blocks for an anchor or query.

The server also exposes resources for an index profile, outline, and chunks through doc://{indexId}/profile, doc://{indexId}/outline, and doc://{indexId}/chunks. Format recognition comes from @file-viewer/core/headless, aligning the reported format with Flyfish File Viewer renderer selection.

Setup and configuration

Run the package directly with npx -y @flyfish-dev/file-ai, or install it globally with npm install -g @flyfish-dev/file-ai. A global installation can be started with file-ai --transport stdio. The standard MCP configuration uses npx as the command and passes -y plus the package name as arguments.

For network access, start the process with --transport http, bind it to a host, and provide a port. The documented local example uses port 8765; its MCP endpoint is /mcp under the configured host and port. Development commands use pnpm for installation, building, testing, validation, and local runs.

Tools and capabilities

  • doc_analyze: create and cache a document index from a local file.
  • doc_read: retrieve blocks, chunks, or anchored content.
  • doc_search: search indexed content with source locations.
  • doc_context: retrieve content near an anchor or query.
  • doc_list_formats: report registered formats, extensions, renderer capabilities, and extractor coverage.

Structured extraction covers text, code, Markdown, JSON, PDF text, DOCX, XLSX and CSV, PPTX, EML, and archive manifests. The File AI MCP server recognizes a broader set of registered extensions, but recognition does not guarantee structured text extraction. For renderer-oriented formats such as CAD, 3D, media, images, geospatial files, drawings, ebooks, and some data assets, results may be limited to profiles, metadata, best-effort text, and explicit warnings.

Limitations and notes

The File AI MCP server is read-only and does not change source documents. Format recognition and content extraction are separate layers: a file can have a recognized profile.format while offering limited extracted content. Agents should inspect extraction details and warnings before treating results as complete.

The server runs locally from the command line and supports stdio and streamable HTTP transports.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about File AI

File AI is an MCP server that connects AI agents to local documents through Flyfish File Viewer core. Its main tools analyze and index files, read blocks or anchors, search cached content, retrieve nearby context, and list format and extractor coverage. It is read-only and supports stdio or streamable HTTP transport.

Install it by running `npx -y @flyfish-dev/file-ai`, or install the package globally with `npm install -g @flyfish-dev/file-ai`. A global installation starts with `file-ai --transport stdio`; MCP clients can also invoke the package directly through a stdio configuration.

Run `npx -y @flyfish-dev/file-ai --transport http --host 127.0.0.1 --port 8765`. The MCP endpoint is then `http://127.0.0.1:8765/mcp`. The server also supports stdio when an HTTP endpoint is not needed.

File AI recognizes formats through Flyfish File Viewer core and currently covers 206 registered extensions. Structured extraction is available for text and code, Markdown, JSON, PDF text, DOCX, XLSX, CSV, PPTX, EML, and archive manifests; other recognized formats may return metadata, best-effort text, and warnings.

No. File AI is read-only and does not mutate source documents. Its operations parse files, cache indexes, read extracted content, search cached material, retrieve context, and report format or extraction information.

Related MCP servers

All in File Systems All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewFile AI AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/file-ai?style=directory)](https://allmcps.com/mcp/file-ai)
HTML Embed
<a href="https://allmcps.com/mcp/file-ai"><img src="https://allmcps.com/api/badge/file-ai?style=directory" alt="File AI on AllMCPs" /></a>