Give your AI agents
the power to act.
The industrial-grade execution layer for the agentic economy. One API to securely route agent intents to real-world vendors with built-in financial guardrails and human-in-the-loop consent.
import basecobalt
# 1. Initialize the BaseCobalt client
client = basecobalt.Client(api_key="bc_live_...")
# 2. Pass the LLM's structured intent directly to the API
response = client.execute(
action="book_flight",
parameters={
"origin": "SFO",
"destination": "JFK",
"date": "2026-05-12",
"max_price": 400
},
require_consent=True # Triggers Human-in-the-Loop approval
)
if response.status == "success":
print(f"Flight booked! Confirmation: {response.data.confirmation_code}")
Infrastructure designed for zero liability.
Stop writing custom vendor integrations. We handle the messy reality of the physical world so you can focus on building smarter AI.
Universal Intent API
Pass structured JSON intents directly from your LLM. We automatically map, normalize, and route the action to thousands of travel and dining vendors.
Human-in-the-Loop Consent
Never spend user funds blindly. When an agent initiates a paid action, we automatically pause execution and fire a webhook requesting explicit user approval.
Ephemeral Virtual Cards
Powered by Stripe Issuing. We generate exact-amount, single-use virtual credit cards for every approved transaction. Zero chargeback risk or runaway AI spending.
Hybrid Execution Engine
We use deterministic official APIs (like Duffel) when available. For legacy vendors, requests fallback seamlessly to our AI-native browser automation layer.
Ready to move beyond chat?
Join hundreds of developers giving their AI agents real-world capabilities.