Your agents remember why, not just what.

Tortoise is a computational memory graph. Every relationship expresses how one belief affects another — when new evidence arrives, confidence propagates and your agent’s world model updates.

Your agents make better decisions, waste fewer tokens, and their thinking becomes auditable and collaborative.

Over MCP.

Capture

Agent sessions flow into a semantic graph automatically.

$ tortoise session capture --file chat.json ✓ 47 turns → 12 decisions, 8 claims, 3 hypotheses

Recall

Future agents query the past — and answer from what you decided.

$ tortoise_suggest_entry_points("database architecture") → "Last session you decided FalkorDB Cloud (confidence 0.9)"

Reason

Belief propagation surfaces what's true — and what's been contradicted.

IMPL a → b confidence: 0.87 NAND c → a confidence: 0.72

Connect your agent →

5,000 Points free. No credit card. Start building in minutes.

1 · Sign up to get your API key, then either:

A · Connect your agent (MCP)

{
  "mcpServers": {
    "tortoise": {
      "command": "python3",
      "args": ["-m", "tortoise.mcp_server"],
      "env": {
        "TORTOISE_API_KEY": "tt_YOUR_KEY",
        "TORTOISE_API_URL": "https://api.premiselabs.co"
      }
    }
  }
}

Paste into .mcp.json (Claude Code), .cursor/mcp.json (Cursor), or claude_desktop_config.json (Claude Desktop). Replace tt_YOUR_KEY with the API key from signup — your agent then reads and writes Tortoise memory automatically.

2 · For scripting and automation, use the CLI:

B · CLI quickstart

$ pip install git+https://github.com/daniel-ospina/tortoise.git $ tortoise init --api-key <key> $ tortoise create-point "Hello world" --kind statement $ tortoise session capture --file chat.json

Self-hosting docs → Docs Login → @premiselabs