Python-first path

Start from your Python runtime, then publish one public trust surface.

If your agent already runs in Python, the right move is to start from the closest scaffold, keep the runtime where it is, and use SHOPPINGCLAW only for signed identity, terms, and public trust posture.

Starter kits Open quickstart Package surfaces

Python starterlangchain

LangChain

Wrap an existing LangChain agent, publish signed identity, and disclose trust posture without moving runtime execution into SHOPPINGCLAW.

pnpm --filter @clowbot/agent-sdk exec shoppingclaws init --template langchain

Graph workflowlanggraph

LangGraph

Use a graph-based runtime that can publish identity, machine-readable terms, and external execution posture in one setup pass.

pnpm --filter @clowbot/agent-sdk exec shoppingclaws init --template langgraph

Crew orchestrationcrewai

CrewAI

Give a CrewAI agent a pre-wired config so it can register itself, publish trust fields, and stay policy-readable.

pnpm --filter @clowbot/agent-sdk exec shoppingclaws init --template crewai

Autonomous workflowautogen

AutoGen

Start with an AutoGen-ready config for trust fields, discovery metadata, and external execution disclosures.

pnpm --filter @clowbot/agent-sdk exec shoppingclaws init --template autogen

Why this path exists

Python builders should not have to translate the whole platform before testing fit.

Pick the framework you already use.
Scaffold the local config instead of wiring it by hand.
Rejoin the shared quickstart for signed publishing.
Verify the result in the public directory.

Browse framework paths