go-docs-mcp
Single-binary server that reads PDFs, Office files, EPUB and images with OCR and search
- Pricing
- Open source
- Platforms
- CLI, macOS, Linux
- Runs
- Locally
- Version
- 1.5.0
- Listing updated
- 3 days ago
About go-docs-mcp
go-docs-mcp is an open-source MCP server for files and documents. It runs locally on your machine and is started by your AI client.
One compiled Go binary exposes 13 tools over stdio, with no Node.js or Python runtime. It reads PDF, text, Markdown, CSV, DOCX, XLSX, PPTX, ODT, EPUB, HTML, RTF, Jupyter notebooks and images (PNG, JPG, WebP, TIFF, BMP). PDF support uses poppler, Office and similar formats use pandoc, and OCR uses tesseract; the format list tool shows which are installed.
Long documents arrive in pieces through offset and max_chars (50,000 characters by default). Text is cached in memory by modification time, and every document is also a docs:// resource that Claude Code can @-mention. Pulled-out PDF images come back as PNG, scaled to 1600 px, up to 10 per call.
The README says everything runs on your machine and no document goes to a cloud API. Access is read-only and confined to one folder, symlinks pointing out of it are refused, and URL downloads (50 MB maximum) refuse private and loopback hosts. It installs with Homebrew, go install or a release binary on macOS, Linux and Windows.
Install
Run in a terminal
claude mcp add go-docs-mcp -- docker run -i --rm ghcr.io/drolosoft/go-docs-mcp:1.5.0Add to claude_desktop_config.json
{
"mcpServers": {
"go-docs-mcp": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/drolosoft/go-docs-mcp:1.5.0"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"go-docs-mcp": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/drolosoft/go-docs-mcp:1.5.0"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"go-docs-mcp": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/drolosoft/go-docs-mcp:1.5.0"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"go-docs-mcp": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/drolosoft/go-docs-mcp:1.5.0"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"go-docs-mcp": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"ghcr.io/drolosoft/go-docs-mcp:1.5.0"
]
}
}
}Configuration
DOCS_MCP_DIRFolder with the documents to serve.RequiredDOCS_MCP_OCR_LANGTesseract language for OCR, for example spa or eng+spa.DOCS_MCP_DISABLE_URLtrue removes the read_url tool.
Tools this server provides (13)
Taken from the project documentation; the running server may differ.
- list_documents
- Lists documents in the configured folder with format, pages and size.
- list_formats
- Lists supported formats and which external tools are installed.
- read_document
- Reads full text, one page or a page range.
- read_url
- Downloads a document from a URL and reads its text.
- convert_to_markdown
- Converts a document to Markdown.
- search_document
- Searches within a document with context.
- get_document_summary
- Gives an overview from the first 3 pages.
- get_document_metadata
- Extracts title, author, dates and page count.
- get_document_outline
- Returns the heading structure.
- extract_tables
- Extracts tables as structured data.
- extract_images
- Extracts PDF images as PNG.
- ocr_document
- Forces OCR on scanned or image-based PDFs.
- read_image
- Extracts text from images with OCR.
What it does
- Many formats
- PDF, Office files, EPUB, HTML, RTF, notebooks and images from one server.
- OCR
- Reads scanned PDFs and images with tesseract.
- Paged output
- offset and max_chars keep long documents from flooding context.
- Tables and images
- Extract tables as structured data and PDF images as PNG.
- Folder-locked and read-only
- Documents are confined to one directory.
- Resource support
- Documents appear as docs:// resources.
Good for
- Summarise a scanned contract with OCR
- Search for a term inside a 300-page PDF
- Pull tables out of a spreadsheet or report
- Convert a document to Markdown
- Read a document from a URL
Pricing
Open source
Open source; free to install
Prices change. Check drolosoft.com before you buy.
go-docs-mcp FAQ
Does it need Node.js or Python?
No. It is a single Go binary, though PDF, Office and OCR support use poppler, pandoc and tesseract.
Does it send documents to the cloud?
The README says no; everything runs locally.
Can it read a URL?
Yes, with read_url, up to 50 MB; private and loopback hosts are refused.
How is access limited?
It is read-only and locked to one folder.
How do I install it?
With brew, go install, or a release download; the registry lists a release binary.
Is go-docs-mcp free?
Yes. go-docs-mcp is open source, so you can use and self-host it for free.
How do I add go-docs-mcp to Claude Code?
Run this command in your terminal: claude mcp add go-docs-mcp -- docker run -i --rm ghcr.io/drolosoft/go-docs-mcp:1.5.0
What are the best alternatives to go-docs-mcp?
The closest alternatives to go-docs-mcp are ASTra, CARTESIAN ECU, Invoices, SigPen and ConvertAudioToText.
Alternatives to go-docs-mcp

ASTra
MCP serverLet agents read, edit, validate and undo Word .docx files with exact operations
Runs locally- 5 tools
- Freemium
- Files and documents

CARTESIAN ECU
MCP serverCheck and repair ECU and TCU flash-file checksums from Claude (Ultimate plan)
Not checked yet- 5 tools
- Remote
- Paid
- Files and documents
Invoices
MCP serverMCP server that reads supplier invoices and returns data plus DATEV, CSV or Excel files
- 4 tools
- Remote
- Paid
- Files and documents

SigPen
MCP serverMCP server that drafts e-signature documents while you approve every send
Not checked yet- 13 tools
- Remote
- Freemium
- Files and documents

ConvertAudioToText
MCP serverTranscribe audio and video links from Claude or ChatGPT and read the results
Not checked yet- 5 tools
- Remote
- Freemium
- Design and media

WhoScribe
MCP serverSearch and read your WhoScribe audio and video transcripts from ChatGPT, read-only
Not checked yet- Remote
- Freemium
- Design and media