fleet-memory/hindsight-docs/docs/sdks/mcp.md
2025-11-27 16:22:16 +01:00

1.9 KiB

sidebar_position
4

MCP Server

Model Context Protocol server for AI assistants like Claude Desktop.

Installation

cd hindsight-cli && cargo build --release

Claude Desktop Setup

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "hindsight": {
      "command": "/path/to/hindsight",
      "args": ["mcp-server"],
      "env": {
        "HINDSIGHT_API_URL": "http://localhost:8888",
        "HINDSIGHT_AGENT_ID": "claude-agent"
      }
    }
  }
}

Environment Variables

Variable Description Default
HINDSIGHT_API_URL Hindsight API URL http://localhost:8888
HINDSIGHT_AGENT_ID Default bank ID Required

Available Tools

Search memories:

{
  "name": "hindsight_search",
  "arguments": {
    "query": "What does Alice do for work?",
    "top_k": 5
  }
}

hindsight_think

Generate response using memories:

{
  "name": "hindsight_think",
  "arguments": {
    "query": "What should I recommend to Alice?"
  }
}

hindsight_store

Store new memory:

{
  "name": "hindsight_store",
  "arguments": {
    "content": "User prefers Python for data analysis",
    "context": "programming discussion"
  }
}

hindsight_agents

List available memory banks:

{
  "name": "hindsight_agents",
  "arguments": {}
}

Usage Example

Once configured, Claude can use Hindsight naturally:

User: "Remember that I prefer morning meetings"

Claude: Uses hindsight_store

"I've noted that you prefer morning meetings."


User: "What do you know about my preferences?"

Claude: Uses hindsight_search

"Based on our conversations, you prefer morning meetings and like Python for data analysis."

Testing

Run standalone:

hindsight mcp-server

Debug mode:

RUST_LOG=debug hindsight mcp-server