← All docs

Sanction + CrewAI (Python)

Route CrewAI agents' LLM calls through Sanction's gateway so every token is metered and capped, and give the crew an authorize tool it must call before any spend. CrewAI uses LiteLLM under the hood, so the integration is a base-URL swap plus a tool.

Prerequisites: a Sanction wallet and agent key (pxy_…). See the Quickstart.

1. Install

pip install crewai httpx

2. Gateway base-URL swap

Point CrewAI's LLM at Sanction's gateway with extra headers.

# sanction_crew.py
import os
from crewai import LLM

SANCTION_KEY = os.environ["SANCTION_AGENT_KEY"]  # pxy_...

# Claude via Sanction gateway
llm = LLM(
    model="anthropic/claude-sonnet-4-20250514",
    api_base="https://getsanction.com/api/gateway/anthropic",
    extra_headers={"x-sanction-key": SANCTION_KEY},
)

# OpenAI via Sanction gateway
llm_openai = LLM(
    model="openai/gpt-4o",
    api_base="https://getsanction.com/api/gateway/openai/v1",
    extra_headers={"x-sanction-key": SANCTION_KEY},
)

Every call through the gateway is automatically metered in your Sanction dashboard.

3. Pre-spend authorization

Before your agent performs any financial action, call POST /api/v1/authorize and honor the result.

# authorize.py
import os, httpx

SANCTION_API = "https://getsanction.com/api/v1"

def authorize(
    action: str,          # "purchase" | "subscribe" | "transfer"
    amount_usd: float,
    merchant: str,
    category: str,
    description: str = "",
) -> dict:
    """Pre-spend check. Returns a dict with 'status': approved | denied | escalated."""
    resp = httpx.post(
        f"{SANCTION_API}/authorize",
        headers={
            "Content-Type": "application/json",
            "x-api-key": os.environ["SANCTION_AGENT_KEY"],
        },
        json={
            "action": action,
            "amount_usd": amount_usd,
            "merchant": merchant,
            "category": category,
            "description": description,
        },
    )
    resp.raise_for_status()
    return resp.json()

4. Putting it together

Expose authorize as a CrewAI tool so the agent must clear spend before it acts.

# crew_agent.py
from crewai import Agent, Task, Crew
from crewai.tools import tool
from sanction_crew import llm
from authorize import authorize

@tool("authorize_spend")
def authorize_spend(action: str, amount_usd: float, merchant: str, category: str, description: str = "") -> str:
    """Check Sanction policy before any financial action."""
    decision = authorize(action, amount_usd, merchant, category, description)
    status = decision["status"]
    if status == "approved":
        return f"APPROVED: proceed with {action} of ${amount_usd} at {merchant}"
    elif status == "escalated":
        return f"ESCALATED: waiting for human approval -- {decision.get('reason', '')}"
    else:
        return f"DENIED: {decision.get('reason', '')} | {decision.get('remediation', '')}"

researcher = Agent(
    role="Procurement Researcher",
    goal="Find and recommend software tools within budget",
    backstory="You evaluate SaaS tools and check spend approval before purchasing.",
    tools=[authorize_spend],
    llm=llm,
)

task = Task(
    description="Find a CI/CD tool under $50/mo. Authorize the purchase via Sanction before proceeding.",
    expected_output="Tool recommendation with Sanction approval status",
    agent=researcher,
)

crew = Crew(agents=[researcher], tasks=[task], verbose=True)

if __name__ == "__main__":
    print(crew.kickoff())

Environment variables

export SANCTION_AGENT_KEY=pxy_live_...    # from POST /agents
export ANTHROPIC_API_KEY=sk-ant-...       # your provider key (forwarded through the gateway)

Reference

Start free →