How to Connect

abrenunt.io exposes its document library via the Model Context Protocol (MCP). Any MCP-compatible AI client can connect and search the collection.

Claude Desktop

Open your Claude Desktop configuration file and add:

{
  "mcpServers": {
    "abrenuntio": {
      "type": "sse",
      "url": "https://abrenunt.io/sse"
    }
  }
}

On macOS, this file is at ~/Library/Application Support/Claude/claude_desktop_config.json. On Windows, it's at %APPDATA%\Claude\claude_desktop_config.json.

Claude Code

Similar to Desktop, add to your project's .mcp.json or global MCP settings:

{
  "mcpServers": {
    "abrenuntio": {
      "type": "sse",
      "url": "https://abrenunt.io/sse"
    }
  }
}

Other MCP clients

SSE endpoint: https://abrenunt.io/sse

Streamable HTTP: https://abrenunt.io/mcp

Available tools

Once connected, your AI assistant will have access to these tools:

Tool	Description
`search_documents`	Hybrid semantic + keyword search across all documents. Supports filtering by document, tags, and authors.
`read_document_part`	Read a specific section or page of a document by section path or page number.
`list_document_parts`	Show the table of contents for a document.
`list_documents`	List available documents, with optional filters by author, tag, or search query.
`get_document_info`	Get metadata about a specific document (author, year, language, etc.).

Free options

Don't have a paid AI subscription? You can still give abrenunt.io a spin:

Claude.ai (free tier) — the easiest option. Go to Settings → Integrations, add https://abrenunt.io/sse as a remote MCP server. No install required, works entirely in the browser.
Cursor or Windsurf (free tiers) — code editors with MCP support. Add the SSE URL to their MCP settings.

A note on result quality

abrenunt.io handles search and retrieval — the quality of the final answer also depends on the AI model you connect it to. Key factors:

Reasoning ability — stronger models synthesise retrieved passages more accurately and are less likely to misread or confabulate.
Context window — larger context windows allow the model to read more retrieved text before answering. Useful for complex questions drawing on multiple documents.
Instruction following — models that reliably follow instructions will make better use of the MCP tools (searching, reading sections, cross-referencing).

In practice, (paid) models such as Claude Sonnet and Claude Opus give noticeably better results than free-tier models for detailed questions. For simple lookups and quick references, free tiers work well.

Fair use

This is a free service. Rate limits are in place. (Don't worry, normal research use should never come close to these limits.)