AgentGuard

What it does

Four runtime guards. Wrap them around your agent. That's the whole API.

Budget guard

Stop an agent before it burns through $X.

from agentguard47 import Guard
with Guard(budget_usd=5.00):
    agent.run(task)

Loop guard

Kill runaway agents before they tool-call 10,000 times.

with Guard(max_tool_calls=50):
    agent.run(task)

Timeout guard

Hard time ceiling on agent execution.

with Guard(timeout_s=300):
    agent.run(task)

Rate guard

Cap tool calls per minute.

with Guard(max_tool_calls_per_min=10):
    agent.run(task)

Open source, Pro, Team

The open source SDK is always free. Pro and Team add the hosted dashboard, longer event history, and email alerts.

	Open source	Pro	Team
Price	Free	$39/mo	$79/mo
Runtime guards	✓	✓	✓
Local telemetry	✓	✓	✓
MIT license	✓	✓	✓
Hosted dashboard	—	✓	✓
Event history	—	500K	5M
Users	1	1	10
Email alerts	—	✓	✓
Team visibility	—	—	✓
	`pip install agentguard47`	Start trial	Start trial

Pro and Team links use Stripe hosted checkout via NEXT_PUBLIC_STRIPE_AGENTGUARD_PRO_LINK / NEXT_PUBLIC_STRIPE_AGENTGUARD_TEAM_LINK env vars. Set these in Vercel before shipping.

Why this exists

I wrote AgentGuard because the existing options all point at a different problem. Lakera is about prompt injection. Guardrails AI is about output validation. Platform-native guardrails ship with a single vendor and lock you in. None of them stop an agent from burning $200 of OpenAI credit in a runaway loop at 2 AM.

AgentGuard is runtime only. It sits between your code and the model, counts cost and tool calls and wall-clock time, and raises before you get the surprise bill. It works the same whether you're calling GPT-5, Claude, a local llama.cpp server, or something I haven't heard of yet.

I built it for my own agents. I run it on my autotrader and on the agents that post this blog. If it works for me at 2 AM, it should work for you.

FAQ

What is AgentGuard?: A small Python SDK you wrap around any agent run. It enforces budget, tool-call, timeout, and rate limits at runtime. If the agent tries to exceed one, AgentGuard stops the run and tells you why.
Why not use OpenAI Agents SDK guardrails?: Because AgentGuard works with every framework, not just OpenAI's. I use it with LangGraph, Claude agents, LlamaIndex, my own scrappy loops, and the Agents SDK. One guardrail layer for every agent you ship.
Can I self-host the dashboard?: The dashboard is Pro-only and hosted by me. If self-hosting is a hard requirement, open an issue on GitHub and we'll talk. The SDK itself is MIT and runs anywhere.
Do you store my agent inputs or outputs?: No. Pro sends aggregate runtime metrics (cost, tool calls, duration) to the hosted dashboard, not prompts or responses. Local telemetry is opt-in and stays on your machine.
Who's behind this?: One developer in Tennessee. Me. Read the about page if you want the long version.