Use the docs with your AI assistant

The ABBYY documentation MCP server lets an AI assistant search and read this site while you work: every product documented here (ABBYY Vantage, Vantage Self-Hosted, FineParser, FineReader Engine, and FlexiCapture) plus the Vantage OpenAPI specifications. Connect it once and your assistant answers from the current docs rather than from training data. The server is read-only and public. It does not connect to a Vantage tenant, process documents, or need ABBYY credentials, and nothing runs on your machine.

This page is about searching the documentation. It is not the Vantage MCP server that connects an assistant to your Vantage tenant to run skills and process documents.

Time to complete: ~5 minutes

Connect your assistant

The server address is the same everywhere: https://docs.abbyy.com/mcp, HTTP transport, no authentication. Pick your client.

Claude Code
VS Code (Copilot)
Cursor
Codex
Claude app

Run this in your terminal:

claude mcp add abbyy-docs https://docs.abbyy.com/mcp --transport http

By default the server is added at the local scope: available to you in the current project only. Use --scope to change that.

Scope	Where it is stored	When to use
`local` (default)	Your Claude Code user configuration, keyed to the current project	Try it out in one project
`project`	`.mcp.json` in the project root, committed with the code	Share the server with everyone who works on the repository
`user`	Your Claude Code user configuration, for all projects	Have it available everywhere on this machine

To share the server through version control, create or edit .mcp.json in the project root instead of running the command:

{
  "mcpServers": {
    "abbyy-docs": {
      "type": "http",
      "url": "https://docs.abbyy.com/mcp"
    }
  }
}

Claude Code reads this file on startup and asks you to approve the server the first time it is used in that project.Restart Claude Code and run /mcp. You should see abbyy-docs with a connected status.

Requires VS Code 1.99 or later with the GitHub Copilot extension signed in.

Open the Command Palette with Cmd+Shift+P (macOS) or Ctrl+Shift+P (Windows/Linux)
Run MCP: Add Server…
Select HTTP as the transport type
Enter the URL https://docs.abbyy.com/mcp
Enter the name abbyy-docs
Choose Workspace (.vscode/mcp.json, shared with the repository) or User (all projects on this machine)

To configure by hand, create or edit .vscode/mcp.json, or run MCP: Open User Configuration for the user-level file:

{
  "servers": {
    "abbyy-docs": {
      "type": "http",
      "url": "https://docs.abbyy.com/mcp"
    }
  }
}

Run MCP: List Servers and confirm abbyy-docs shows as connected. In Copilot Chat, switch to Agent mode; MCP tools are only available there.

Open Cursor Settings and select MCP
Click Add new MCP server
Enter the name abbyy-docs and the URL https://docs.abbyy.com/mcp

Or add the entry to .cursor/mcp.json in the project (shared with the repository) or ~/.cursor/mcp.json (all projects):

{
  "mcpServers": {
    "abbyy-docs": {
      "url": "https://docs.abbyy.com/mcp"
    }
  }
}

The server appears in the MCP settings list with a green status once Cursor has connected.

Run this in your terminal:

codex mcp add abbyy-docs --url https://docs.abbyy.com/mcp

Restart Codex. The server is available in every project on this machine.

Open Settings and select Connectors
Click Add custom connector
Enter the name ABBYY docs and the URL https://docs.abbyy.com/mcp
Click Add

There is no sign-in step because the server needs no authentication. In a chat, open the attachments menu (the plus icon) and enable the connector.

Verify the connection

Ask your assistant a question about any ABBYY product:

“What parameters does the Vantage Processing API require to submit a document?”

The assistant calls search_abbyy_documentation, and usually follows up with query_docs_filesystem_abbyy_documentation to read the full page before answering. In clients that show tool calls, those are the names to look for.

If the server does not appear in your client, check that the URL is entered exactly as https://docs.abbyy.com/mcp with no trailing slash, and that your machine has network access to reach it.

What your assistant can do

The server exposes three tools. The assistant decides when to call each one based on your question.

Tool	What it does
`search_abbyy_documentation`	Semantic search across the whole site. Returns relevant passages with page titles and links. Accepts an optional `language` filter (`en`, `de`, `es`, `fr`, `ja`); results default to English
`query_docs_filesystem_abbyy_documentation`	Reads the documentation as a read-only virtual filesystem. The assistant uses it to `cat` or `head` full pages, `rg` for exact keywords, and query the OpenAPI specs with `jq`. Nothing runs on your machine
`submit_feedback`	Reports a documentation problem to the ABBYY docs team, with the page path and a description of what is wrong

Example questions:

“How do I configure a Document Skill in Vantage?”
“What does the minConfidence parameter do?”
“How do I set up an LLM connection in Vantage?”
“Which outputType values does FineParser accept?”
“List the endpoints in the Vantage Reporting API”

If your assistant finds a page that is wrong or out of date, ask it to report the problem. It calls submit_feedback with the page and the issue, which lands directly with the docs team.

Tips for effective questions

Be specific. Include the feature name or API endpoint rather than asking generally
Name the product. The server covers several products, so “Vantage Process Skill” or “FineParser fast mode” narrows the search
Use product terminology. Terms like “Process Skill”, “Document Skill”, “Skill Designer”, and “tenant” return more precise results
Ask for another language. Ask in German, Spanish, French, or Japanese, or say “search the German docs”, and the assistant applies the matching language filter
Ask follow-up questions. Your assistant retains context, so you can narrow results with follow-up prompts
Combine with your code. Paste an API response or config snippet and ask the assistant to explain it using the docs

Troubleshooting

The server does not appear in my client

Check that the configuration was saved and, for JSON files, that there are no trailing commas. Restart the client fully. In a project with a .mcp.json, approve the server when Claude Code prompts you. In VS Code, MCP support needs version 1.99 or later; check Help → About.

The tools are never called

Assistants call MCP tools based on relevance. Ask more explicitly (“Search the ABBYY docs for…”) or name a specific product feature. In Copilot Chat, make sure the mode selector is set to Agent; the tools are unavailable in Ask and Edit modes.

Search returns irrelevant results

Rephrase using terms from the product interface or documentation. Avoid generic terms like “workflow” or “settings” on their own; pair them with the product name or a specific feature.

Results are in the wrong language

Search defaults to English. Ask the assistant to search the docs in the language you want, and it sets the language filter.

Remove the server

Claude Code
VS Code (Copilot)
Cursor
Codex
Claude app

claude mcp remove abbyy-docs

Add --scope project or --scope user to match the scope you added it with.

Run MCP: List Servers, select abbyy-docs, and choose Remove. Or delete the entry from .vscode/mcp.json or your user MCP configuration.

Delete the server from Cursor Settings → MCP, or remove the entry from .cursor/mcp.json or ~/.cursor/mcp.json.

codex mcp remove abbyy-docs

If you added the server under an earlier name such as abbyy-vantage, it keeps working. Remove it the same way, using that name.

Next steps

Getting Started with the Vantage API

Make your first call and authenticate with OAuth 2.0

LLMs in ABBYY Vantage

Bring your own LLM connection or use ABBYY Phoenix Plus

Working with skills

Choose, build, and publish the skills that process your documents

FineParser and AI coding agents

Use the same server to write a FineParser integration

​Connect your assistant

​Verify the connection

​What your assistant can do

​Tips for effective questions

​Troubleshooting

​Remove the server

​Next steps

Getting Started with the Vantage API

LLMs in ABBYY Vantage

Working with skills

FineParser and AI coding agents

Connect your assistant

Verify the connection

What your assistant can do

Tips for effective questions

Troubleshooting

Remove the server

Next steps