Now in Beta for Early Adopters

Stop burning cash on
invisible AI agents.

The observability layer for autonomous spend. Track every model call, set hard limits, and attribute costs down to the specific agent workflow.

Start Monitoring FreeBook a Demo
dashboard.keel.dev

TOTAL AGENT BURN

$1,240.89

12% lower than last week

ACTIVE AGENTS

47

BLOCKED OVERAGES

12

Last triggered: 2 mins ago

import anthropic
import keel

client = keel.wrap(
    anthropic.Anthropic(),
    agent="document-pipeline",
    api_url="https://api.keel.dev",
    api_key="keel_sk_live_...",
)

# Every call is tracked automatically
response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": prompt}],
)

Monitor your usage intuitively

Real-time spend breakdowns per agent, model, and function — all in one place.

Easy to import

One line to wrap your existing Anthropic client. No refactoring required.

Configurable YAML for full control

Set budgets, alerts, and rate limits per agent with a simple config file.

keel.yaml
agents:
  document-pipeline:
    budget_usd: 50.00        # hard cap per month
    alert_at: 0.80           # alert at 80% usage
    model_preference: sonnet # fallback if opus exceeds budget

  support-bot:
    budget_usd: 10.00
    alert_at: 0.90
    rate_limit: 100          # max calls per hour

  data-extractor:
    budget_usd: 5.00
    alert_at: 0.75
    block_on_exceed: true    # hard stop, no overages

Infrastructure for the Agent Economy

You wouldn't run a team without a CFO. Don't run agents without Keel.

Real-Time Attribution

Trace every API call back to the specific agent and workflow. No more mystery line items on your Stripe bill.

Programmable Limits

Set hard budget caps per agent. If an infinite loop starts burning cash, we kill the process instantly.

One Integration

Works with LangChain, AutoGPT, and custom stacks. Drop in our SDK and get visibility in minutes.