Talon MCP Server

MCP (Model Context Protocol) server for the Talon multi-model data engine. Enables AI agents in Cursor, Claude Desktop, and other MCP-compatible clients to directly query and operate all 9 Talon engines.

Supported Engines & Tools

| Engine | Tools | Description | |--------|-------|-------------| | SQL | talon_sql_query, talon_sql_execute | Full SQL with vector/geo functions | | KV | talon_kv_get/set/delete/scan/mget/incr | Redis-compatible key-value store | | Vector | talon_vector_search/insert/create_index | HNSW similarity search | | TimeSeries | talon_ts_query/write/create | Time-series data with aggregation | | MessageQueue | talon_mq_publish/poll/ack/create_topic/list_topics | Consumer groups, DLQ, priority | | FTS | talon_fts_search/index_doc/hybrid_search/create_index | BM25 + hybrid (BM25+Vector RRF) | | GEO | talon_geo_search/add/create | Geospatial radius/box search | | Graph | talon_graph_query/shortest_path/add_vertex/add_edge/create | Property graph with BFS, shortest path | | AI | talon_ai_session, talon_ai_memory | Session/Context/Memory management | | Admin | talon_list_tables/describe_table/server_info/persist/raw_execute | Schema inspection, server status, raw commands |

Total: 35+ tools covering all Talon capabilities.

Quick Start

1. Build

cd talon-mcp
npm install
npm run build

2. Start Talon Server

./talon-server --http-port 8080 --data ./data

3. Configure in Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "talon": {
      "command": "node",
      "args": ["/Users/dark/WebstormProjects/talon-mcp/dist/index.js"],
      "env": {
        "TALON_URL": "http://localhost:8080"
      }
    }
  }
}

4. Configure in Claude Desktop

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

{
  "mcpServers": {
    "talon": {
      "command": "node",
      "args": ["/Users/dark/WebstormProjects/talon-mcp/dist/index.js"],
      "env": {
        "TALON_URL": "http://localhost:8080"
      }
    }
  }
}

Environment Variables

| Variable | Default | Description | |----------|---------|-------------| | TALON_URL | http://localhost:8080 | Talon server HTTP endpoint | | TALON_TIMEOUT | 30000 | Request timeout in milliseconds |

Examples

Once configured, AI agents can interact with Talon directly:

"Show me all tables in the database"
→ calls talon_list_tables

"Create a users table with id, name, and a 384-dim embedding column"
→ calls talon_sql_execute with CREATE TABLE users (id INTEGER PRIMARY KEY, name TEXT, emb VECTOR(384))

"Search for vectors similar to [0.1, 0.2, ...] in the embeddings index"
→ calls talon_vector_search

"Set cache key session:abc to token123 with 1 hour TTL"
→ calls talon_kv_set with ttl=3600

"Find articles about 'machine learning' using hybrid search"
→ calls talon_fts_hybrid_search combining BM25 + vector similarity

Development

npm run dev    # Watch mode with tsx
npm run build  # Production build
npm start      # Run built server

License

MIT