Governance for AI Agents
Stop unwanted agent actions before they run.
Every tool call your AI agent makes — checked against your policies before execution. Not after. Not sometimes. Every single one.
$ pip install agentbouncr$ npm install agentbouncrNot on the list? Not getting in.
Everything between your agents and production — from first tool call to compliance report.
Agent Registry
Every AI agent in your organization: who built it, what tools it can access, when it was last active, and how to shut it down. One registry, full visibility.
Block Before It Runs
Compromised agent detected? Every action blocked before it executes. Board-level resilience.
EU AI Act Evidence
Built in Europe. Produces the audit evidence for AI-Act obligations — risk categorization, human-oversight records, full traceability. Evidence support, not a compliance guarantee. Enforcement begins August 2026.
Policy Engine
Every tool call checked before execution. Define rules once in code or via dashboard, enforce everywhere. No policy, no entry.
Kill Switch
One click. Every further action is blocked. Every affected tool call logged. On-demand incident response.
Tamper-Evident Audit Trail
Cryptographic hash-chain on every action. Append-only, verifiable on demand, exportable, auditor-ready. Hand it to your compliance team — they’ll thank you.
See it in action
Monitor all your agents, policies, and audit events from a single dashboard.
What happens when your agent goes rogue?
Autonomous agents make autonomous decisions. Agent governance is the question of who controls them — and when.
Without AgentBouncr
Your sales agent autonomously decides to export the entire customer database.
The export runs. No one notices. No log. No alert. Compliance violation — discovered weeks later in an audit.
With AgentBouncr
Same agent, same decision. AgentBouncr checks the policy.
export_customer_list → DENIED. Action blocked before execution. Event logged. Team alerted. Zero data leaked.
Agent governance in three steps
From policy to enforcement to audit — agent governance for your AI agents in minutes, not weeks.
Define Policies
Define policies in code or via dashboard. Works with MCP-compatible tools out of the box. Templates for common patterns included.
Connect Your Agents
Connect in 3 lines of code. Python SDK, TypeScript SDK, or REST API. Native LangChain and CrewAI integrations (Python). Works with any agent that makes tool calls via HTTP/REST.
Monitor & Respond
Live audit dashboard, automated alerts, and a kill-switch for on-demand incident response.
Ready to secure your agents?
Built for teams shipping AI agents to production
From a single LangChain agent to a fleet of autonomous systems — AgentBouncr scales with your governance needs.
LangChain Agent with Tool Governance
Your LangChain agent calls 12 different tools. AgentBouncr checks every call against your policies. Block sensitive operations, require approval for high-risk actions, log everything.
Multi-Agent Oversight
5 teams, 20 agents, 50 tools. One dashboard. See who’s doing what, stop any agent on demand, hand the audit trail to your compliance team.
EU AI Act Evidence
Article 14 requires human oversight of AI systems. AgentBouncr supports it: policy-based control, kill switch, and tamper-evident records of every reported decision — evidence support, not a compliance guarantee.
Free during Alpha
All features included. No credit card required. Sign up and start governing your AI agents.
Everything you need to govern AI agents — included from day one.
- Unlimited agents
- Unlimited evaluations
- Kill switch
- Custom policies + builder
- Approval workflows
- Full audit trail
- Webhooks + CSV export
- PDF compliance reports
- Multi-tenant workspaces
No credit card required. Pricing will be introduced after alpha.
Frequently asked questions
Can't find what you're looking for? Reach out to our team and we'll get back to you.
What is AgentBouncr?
AgentBouncr is an open-core governance framework that sits between your AI agents and the tools they use. It enforces policies, logs every action, and gives you a kill-switch to block further agent actions on demand.
How does the policy engine work?
You define policies via the dashboard or programmatically via the REST API. Specify which tools each agent can access — allow, deny, or require human approval. Policies are evaluated on every tool call before execution. No code deployment needed to change a policy.
Is my audit data tamper-evident?
Yes. Every audit event is linked via a SHA-256 hash chain. Any modification to historical records breaks the chain — you can verify integrity on demand via the audit verification endpoint.
Can I self-host AgentBouncr?
Absolutely. The core is source-available under ELv2. The Enterprise edition adds multi-tenant isolation, OIDC/SAML SSO, Stripe billing, and a managed dashboard.
What AI frameworks are supported?
AgentBouncr is framework-agnostic. It ships native integrations for LangChain and CrewAI (Python) and works with any agent that makes tool calls via HTTP — REST API included.
How does the free tier work?
AgentBouncr is currently in Open Alpha. All features are included for free — unlimited agents, unlimited evaluations, full audit trail, kill switch, and multi-tenant workspaces. No credit card required. No limits. We’ll introduce pricing after the alpha phase.