Control your AI agents.
Set permissions, cap spending, and log every action across your agent fleet — in minutes, not a compliance project.
- One HTTP call, any framework
- Dollar caps per agent
- Kill switch for every bot
- Free for up to 3 agents
Live in Under an Hour
Connect your agents, define the rules, and let AgentGovern enforce them.
Connect Your Agents
Point your AI agents at AgentGovern with one HTTP call before each tool use. Every agent gets its own scoped API key.
Define Permissions
Pick which tools, APIs, and data each agent can touch, and give each one a dollar cap.
Watch and Review
Watch every call land in a live feed, and pull the full log whenever you need to know what happened.
// before the agent touches anything real
const d = await agentgovern.evaluate({
action_type: "refund",
resource_type: "api",
resource: "stripe.refunds.create",
cost_cents: 4900,
});
if (d.decision !== "allowed") return d; // blocked or waiting on a humanYou stay in control
The three things you need once your bots touch real money and real systems — and nothing you don't.
Permissions per tool
Say exactly what each agent may do: allow, deny, or ask you first — per tool, API, and data source. Wildcards like github.* keep it short. Anything you didn't allow is blocked.
Spending caps per agent
Give every bot a budget. A call that would go over is blocked before it runs — no surprise bills.
Kill switch + append-only log
Pause any agent and every call it makes is blocked, instantly. Every decision is written to a log that can't be edited — not even by you.
You approve the risky stuff
Refunds, deletes, shell commands, anything over $50 — send them to your approval queue. The agent waits until you say yes.
Know when something's off
Blocked calls, a bot creeping toward its cap, a sudden burst of requests — all in a live feed, and in Slack if you want it.
Audit trail and one-click export
Every request and decision, filterable by agent, date, and action. Export CSV or JSON when a customer — or future you — asks what happened.
Free for your first 3 agents
Free
For your first few bots.
- Up to 3 agents
- Unlimited policy evaluations
- Allow / deny / approval rules with wildcards
- Per-agent spending caps
- Human-in-the-loop approval queue
- Live alert feed
- Audit export (last 7 days)
Pro
productionEverything you need to run your fleet in production.
- Unlimited agents
- Everything in Free
- Slack alerts for blocks, spend, and approvals
- Full audit history with CSV & JSON export
- Tunable spend-warning and burst thresholds
Know what every agent is doing, right now
Free to start. No credit card required. Connect your first agent in minutes.