sidclaw 0.2.0

Python SDK for SidClaw — governance for AI agents

sidclaw

Python SDK for SidClaw — governance for AI agents.

Identity. Policy. Approval. Trace. Four primitives that give you control over what your AI agents can do.

Install

pip install sidclaw

Quick Start

from sidclaw import SidClaw

client = SidClaw(api_key="ai_...", agent_id="your-agent-id")

# Evaluate an action against the policy engine
decision = client.evaluate({
    "operation": "send_email",
    "target_integration": "email_service",
    "resource_scope": "outbound_email",
    "data_classification": "confidential",
})

if decision.decision == "allow":
    send_email(...)
    client.record_outcome(decision.trace_id, {"status": "success"})
elif decision.decision == "approval_required":
    approval = client.wait_for_approval(decision.approval_request_id)
    if approval.status == "approved":
        send_email(...)
        client.record_outcome(decision.trace_id, {"status": "success"})
elif decision.decision == "deny":
    print(f"Denied: {decision.reason}")

Governance Decorator

from sidclaw import SidClaw
from sidclaw.middleware import with_governance, GovernanceConfig

client = SidClaw(api_key="ai_...", agent_id="your-agent-id")

@with_governance(client, GovernanceConfig(
    operation="send_email",
    target_integration="email_service",
    data_classification="confidential",
))
def send_email(to: str, subject: str, body: str) -> None:
    email_service.send(to=to, subject=subject, body=body)

send_email("[e-mail]", "Follow-up", "Hello...")
# allow -> executes immediately
# approval_required -> waits for human approval, then executes
# deny -> raises ActionDeniedError

Async

from sidclaw import AsyncSidClaw

async with AsyncSidClaw(api_key="ai_...", agent_id="...") as client:
    decision = await client.evaluate({
        "operation": "send_email",
        "target_integration": "email_service",
        "resource_scope": "outbound_email",
        "data_classification": "confidential",
    })

Framework Integrations

pip install sidclaw[langchain]      # LangChain
pip install sidclaw[crewai]         # CrewAI
pip install sidclaw[openai-agents]  # OpenAI Agents SDK
pip install sidclaw[pydantic-ai]    # Pydantic AI
pip install sidclaw[mcp]            # MCP governance proxy
pip install sidclaw[all]            # Everything

LangChain

from sidclaw.middleware.langchain import govern_tools

governed = govern_tools(my_tools, client=client)
agent = create_react_agent(llm, governed)

CrewAI

from sidclaw.middleware.crewai import govern_crewai_tool

governed_tool = govern_crewai_tool(my_tool, client=client)

OpenAI Agents SDK

from sidclaw.middleware.openai_agents import govern_function_tool

tool, handler = govern_function_tool(tool_def, my_handler, client=async_client)

MCP Governance Proxy

from sidclaw.mcp import GovernanceMCPServer, GovernanceMCPServerConfig

server = GovernanceMCPServer(GovernanceMCPServerConfig(
    api_key="ai_...",
    agent_id="agent-001",
    upstream_command="npx",
    upstream_args=["-y", "@modelcontextprotocol/server-postgres", "postgresql://..."],
))
await server.start()

Or use the CLI:

SIDCLAW_API_KEY=ai_... SIDCLAW_AGENT_ID=agent-001 \
SIDCLAW_UPSTREAM_CMD=npx SIDCLAW_UPSTREAM_ARGS="-y,@modelcontextprotocol/server-postgres,postgresql://..." \
sidclaw-mcp-proxy

Webhook Verification

from sidclaw import verify_webhook_signature

is_valid = verify_webhook_signature(
    payload=request.body,
    signature=request.headers["X-Webhook-Signature"],
    secret="whsec_...",
)

Error Handling

from sidclaw import ActionDeniedError, ApprovalTimeoutError, RateLimitError

try:
    decision = client.evaluate(...)
except ActionDeniedError as e:
    print(f"Denied: {e.reason}, trace: {e.trace_id}")
except RateLimitError as e:
    print(f"Rate limited, retry after {e.retry_after}s")
except ApprovalTimeoutError as e:
    print(f"Approval timed out for {e.approval_request_id}")

Links

• Documentation
• TypeScript SDK
• Dashboard
• GitHub