Gavrun
Gavrun provides an agent action control plane that verifies every AI agent tool call before execution, checking tool access, model use, budget, data risk, and approvals. The platform generates decision IDs for every allow, hold, or block action, and integrates with major agent frameworks including LangChain, CrewAI, and OpenAI Agents SDK. Gavrun is deployable on customer infrastructure and available through AWS Marketplace.
- Artificial Intelligence
- Developer Tools
- AI Agents
- Enterprise Software
- Cybersecurity
- Regulatory & Compliance Technology
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Production AI agents can call tools that move money, delete data, email customers, or spin down servers without oversight. Input guardrails check the prompt but not the actual tool call, leaving no control at the point of action—by the time an agent fires a tool, the refund is sent or the rows are gone.
Solution
Gavrun provides an agent action control plane that decides on every agent call before it runs. The platform checks tool access, model use, budget, data risk, and approval requirements, generating a decision ID for every allow, hold, or block. Safe calls proceed with minimal latency, while risky actions pause for policy enforcement or human review. Gavrun sits in front of existing agent frameworks and runtimes, meaning teams can deploy it without rewriting their agent stack. The system is designed to run on the customer's own infrastructure and is available through AWS Marketplace, giving enterprises full data control and compliance alignment.
Target Audience
Engineering and platform teams at enterprises running production AI agents that interact with sensitive systems—including finance, operations, customer support, and data infrastructure—who need enforceable governance over agent actions.
Features
- Real-time action verification covering tool access, model use, budget limits, data risk, and approvals before execution
- Decision IDs and a unified audit trail for every allow, hold, and block action across all agents
- Native integration with LangChain, CrewAI, LlamaIndex, OpenAI Agents SDK, Anthropic Claude, Pydantic AI, n8n, Microsoft Agent Fx, Google ADK, and Amazon AgentCore
- Self-hosted deployment option with AWS Marketplace availability
- Policy-based action gating with human-in-the-loop approval workflows for high-risk operations