Skip to main content
Glama

Arke Institute MCP Server

Official
by Arke-Institute
GitHub
TypeScript
OverviewInspectNewEndpointsSchemaRelated ServersReviewsScore
Need Help?View Source CodeReport Issue

Arke Institute MCP Server

A Model Context Protocol (MCP) server that provides AI assistants with semantic search capabilities across the Arke Institute's extensive archive of NARA (National Archives and Records Administration) records and presidential libraries.

Features

  • Semantic Search: Natural language queries powered by OpenAI embeddings and Pinecone vector search

  • Rich Entity Types: Search across institutions, collections, series, file units, and digitized objects

  • Extracted Text: Access OCR'd content from scanned documents and PDFs

  • Complete Metadata: Full NARA catalog records, access restrictions, physical locations, and hierarchical relationships

  • Fast Responses: Sub-second search times with parallel API processing

  • Easy Integration: Works with Claude Desktop, Cloudflare AI Playground, and any MCP client

What is Arke Institute?

The Arke Institute provides semantic access to historical archives, starting with digitizing and indexing the complete holdings of the National Archives. The search API powers discovery across millions of historical documents, photographs, and records.

Installation

Deploy to Cloudflare Workers

Deploy to Workers

Or via command line:

npm create cloudflare@latest -- arke-mcp-server --template=arke-institute/arke-mcp cd arke-mcp-server npm run deploy

Your MCP server will be deployed to: arke-mcp-server.<your-account>.workers.dev/sse

Local Development

git clone https://github.com/arke-institute/arke-mcp.git cd arke-mcp/arke-mcp-server npm install npm run dev

The server runs at http://localhost:8787

MCP Tool Reference

search_arke

Perform semantic search across Arke Institute archives.

Parameters:

  • query (string, required): Natural language search query

  • topK (number, optional): Number of results (1-100, default: 10)

  • namespaces (array, optional): Filter by entity type(s)

Available Namespaces:

  • institution - Institutional collections

  • collection - Record collections

  • series - Record series

  • fileUnit - File units

  • digitalObject - Digital objects (scanned documents, images, PDFs with extracted text)

Returns:

Formatted search results including:

  • Similarity scores (0-1 range, higher = better match)

  • Entity titles and descriptions

  • NARA identifiers and persistent identifiers (PIs)

  • Date ranges and record types

  • Parent/child entity relationships

  • Physical locations and access restrictions

  • IPFS content identifiers (CIDs)

  • Extracted text from digitized documents

Usage Examples

Example 1: General Search

{ "query": "Apollo 11 moon landing mission", "topK": 5 }

Searches all entity types for Apollo 11 content.

Example 2: Search Digitized Documents

{ "query": "World War II photographs", "topK": 10, "namespaces": ["digitalObject"] }

Searches only digitized objects (with extracted text) for WWII photos.

Example 3: Search File Units

{ "query": "presidential speeches on climate change", "topK": 20, "namespaces": ["fileUnit", "digitalObject"] }

Searches file units and digitized objects for climate-related presidential speeches.

Connect to Claude Desktop

To use this MCP server with Claude Desktop, add it to your configuration:

  1. Open Claude Desktop Settings > Developer > Edit Config

  2. Add the server configuration:

{ "mcpServers": { "arke": { "command": "npx", "args": [ "mcp-remote", "https://arke-mcp-server.<your-account>.workers.dev/sse" ] } } }

For local development:

{ "mcpServers": { "arke": { "command": "npx", "args": [ "mcp-remote", "http://localhost:8787/sse" ] } } }

  1. Restart Claude Desktop

  2. You should see the search_arke tool available

Connect to Cloudflare AI Playground

  1. Go to https://playground.ai.cloudflare.com/

  2. Enter your deployed MCP server URL: arke-mcp-server.<your-account>.workers.dev/sse

  3. Start using the search_arke tool directly in the playground

Example Conversations

Finding Historical Documents

User: "Find documents about the Space Shuttle Discovery missions"

Claude (using search_arke):

{ "query": "Space Shuttle Discovery missions", "topK": 10 }

Returns relevant file units, digitized speeches, and mission records with extracted text content.

Researching Presidential Libraries

User: "Show me Clinton administration documents about Japan relations in the 1990s"

Claude (using search_arke):

{ "query": "Clinton administration Japan relations 1990s", "topK": 15, "namespaces": ["fileUnit", "digitalObject"] }

Returns presidential library materials with dates, locations, and full text content.

Architecture

┌─────────────┐ │ AI Client │ (Claude Desktop, AI Playground) └──────┬──────┘ │ MCP Protocol (SSE/HTTP) ▼ ┌─────────────────────────────┐ │ Arke MCP Server │ │ (Cloudflare Worker) │ │ │ │ - search_arke tool │ │ - Dynamic namespaces │ │ - Result formatting │ └──────┬──────────────────────┘ │ HTTPS ▼ ┌─────────────────────────────┐ │ Arke Search API │ │ search.arke.institute │ │ │ │ - OpenAI embeddings │ │ - Pinecone vector search │ │ - IPFS content retrieval │ └─────────────────────────────┘

Project Structure

arke-mcp-server/ ├── src/ │ ├── index.ts # MCP server implementation │ ├── types.ts # TypeScript type definitions │ ├── clients/ │ │ └── arke.ts # Arke Search API client │ └── utils/ │ └── formatters.ts # Result formatting for AI ├── package.json ├── wrangler.jsonc # Cloudflare Worker config ├── tsconfig.json └── README.md

API Endpoints

  • /sse - Server-Sent Events endpoint for MCP protocol (recommended)

  • /mcp - Standard HTTP MCP endpoint

  • / - Server information and health check

Development

Type Checking

npm run type-check

Code Formatting

npm run format

Deploy to Production

npm run deploy

Performance

  • Search latency: ~500-900ms (including vector search, entity fetching, and formatting)

  • Namespace fetching: Cached at server initialization

  • Concurrent searches: Fully supported via Cloudflare Workers

Limitations

  • Maximum 100 results per query (topK parameter)

  • Searches are read-only (no write operations)

  • Rate limits apply per Cloudflare Workers free tier (or your plan)

Related Projects

Contributing

Contributions welcome! Please open issues or pull requests on GitHub.

License

MIT License - see LICENSE file for details

Support

For questions or issues:

Acknowledgments

This server cannot be installed

-
security - not tested
F
license - not found
-
quality - not tested

How are these scores calculated?

remote-capable server

The server can be hosted and run remotely because it primarily relies on remote services or has no dependency on the local environment.

Enables semantic search across the Arke Institute's extensive archive of NARA records and presidential libraries using natural language queries. Provides access to millions of historical documents, photographs, and records with OCR'd content and complete metadata.

  1. Features
    1. What is Arke Institute?
      1. Installation
        1. Deploy to Cloudflare Workers
        2. Local Development
      2. MCP Tool Reference
        1. search_arke
      3. Usage Examples
        1. Example 1: General Search
        2. Example 2: Search Digitized Documents
        3. Example 3: Search File Units
      4. Connect to Claude Desktop
        1. Connect to Cloudflare AI Playground
          1. Example Conversations
            1. Finding Historical Documents
            2. Researching Presidential Libraries
          2. Architecture
            1. Project Structure
              1. API Endpoints
                1. Development
                  1. Type Checking
                  2. Code Formatting
                  3. Deploy to Production
                2. Performance
                  1. Limitations
                    1. Related Projects
                      1. Contributing
                        1. License
                          1. Support
                            1. Acknowledgments

                              New MCP Servers

                              MCP directory API

                              We provide all the information about MCP servers via our MCP API.

                              curl -X GET 'https://glama.ai/api/mcp/v1/servers/Arke-Institute/arke-mcp-server'

                              If you have feedback or need assistance with the MCP directory API, please join our Discord server