Mnemo

Mnemo provides MCP access to Gemini context caching, letting your AI assistants load large codebases, docs, PDFs, and more into a fast, cost-efficient contextual cache. You can query against cached content with natural language, benefiting from perfect recall, low latency, and simplified workflows.

How to use

You connect an MCP client to Mnemo to load sources into Gemini’s context cache and then query that cache with natural language. Use an HTTP MCP endpoint when you want a remote, managed interface, or a local server endpoint for development and testing. Mnemo handles loading sources (GitHub repos, docs, PDFs, URLs) and querying them to produce responses without chunking or separate embeddings steps.

How to install

Prerequisites: you need Node-compatible tooling on your machine. You will run a local server for development and then expose an MCP endpoint you can point your client at.

# Clone and install
git clone https://github.com/logos-flux/mnemo
cd mnemo
bun install

# Set your Gemini API key
export GEMINI_API_KEY=your_key_here

# Start the server
bun run dev

Configuration and usage notes

You can load multiple sources into a single cache or separate caches for different projects. Mnemo provides a simple JSON-based MCP configuration to expose an HTTP endpoint that your clients can use to load, query, list, evict, and monitor caches.

MCP config examples (HTTP endpoints)

{
  "mcpServers": {
    "mnemo": {
      "type": "http",
      "url": "http://localhost:8080/mcp"
    }
  }
}

MCP config examples (Cloudflare Worker endpoint)

{
  "mcpServers": {
    "mnemo": {
      "type": "http",
      "url": "https://mnemo.<your-subdomain>.workers.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_AUTH_TOKEN"
      }
    }
  }
}

Usage flow tips

• Load a source once to populate the cache. For example, bring in a GitHub repo, a docs site, or a PDF. Mnemo keeps the tokens cached so you don’t pay per query for the loaded content.
• Query with natural language. Ask questions like “How do I initialize the project?” or “What are the API endpoints for this repo?” and Mnemo will respond using the cached context.
• List, evict, or refresh caches as needed to manage costs and relevance.

Available tools

context_load

Load GitHub repos, URLs, PDFs, or local dirs into Gemini cache

context_query

Query a cached context with natural language

context_list

Show active caches with token counts and expiry

context_evict

Remove a cache

context_stats

Get usage statistics with cost tracking

context_refresh

Reload a cache with fresh content