clawdb 0.1.1

Zero-config database for AI agents.

clawdb (Python SDK)

The official Python client for ClawDB.

This package is a network client, not the database engine itself. It connects to a running clawdb-server instance locally or to a hosted cloud endpoint.

pip install clawdb

Before using the SDK locally, make sure clawdb-server is running. The intended local-first path is:

npx @clawdb/cli@latest init

If you operate your own server, set CLAWDB_ENDPOINT to that address instead.

Sync usage

from clawdb import ClawDB

db = ClawDB(endpoint="http://localhost:50050", agent_id="my-agent")

# Store a memory
id = db.memory.remember("Deploy at 3 PM UTC", memory_type="task", tags=["ops"])

# Search
results = db.memory.search("deploy schedule", top_k=5)
for r in results:
    print(f"{r.score:.3f}  {r.memory.content}")

# Recall by ID
memories = db.memory.recall([id])

# Forget
db.memory.forget(id)

Async usage

import asyncio
from clawdb.aio import AsyncClawDB

async def main():
    async with AsyncClawDB(endpoint="http://localhost:50050") as db:
        id = await db.memory.remember("Async test")
        results = await db.memory.search("test")

asyncio.run(main())

Branches

branch = db.branches.fork("my-experiment")
db.memory.remember("hypothesis", tags=["experiment"])  # writes to current branch
db.branches.merge("my-experiment", into="trunk")

LangChain integration

from clawdb.langchain import ClawDBRetriever
retriever = ClawDBRetriever(db=db, top_k=10)
docs = retriever.get_relevant_documents("deploy schedule")

Configuration

Env var	Description
CLAWDB_ENDPOINT	gRPC endpoint (default http://localhost:50050)
CLAWDB_API_KEY	API key (ck_live_... or ck_test_...)
CLAWDB_AGENT_ID	Agent identifier

Runtime model

• Python talks to clawdb-server over gRPC.
• clawdb-server hosts the Rust runtime and storage engine.
• For cloud usage, point CLAWDB_ENDPOINT at your hosted deployment and provide CLAWDB_API_KEY.

Development

pip install -e ".[dev]"
pytest