Preloop
Preloop is an open-source control plane for AI agents, designed to provide centralized governance, security, and observability for fleets of AI agents. It enables organizations to onboard existing agents such as OpenClaw, Hermes, Preloop Code, Codex CLI, Cursor, Gemini CLI, and any MCP-compatible agent with a single command. Once onboarded, all tool calls are routed through the Preloop MCP Firewall, and model traffic is directed through a central gateway, allowing for human approvals, budget controls, runtime observability, and comprehensive audit trails. The platform addresses the growing need for AI agent governance in enterprise environments. It provides a self-hostable solution that helps organizations meet regulatory requirements such as the EU AI Act, Cyber Resilience Act, NIS2, and DORA. By acting as a middle layer between agents and their tools/models, Preloop enables security teams to enforce policies, approve sensitive actions from mobile or Slack, and maintain searchable evidence of agent activity. Preloop is positioned as an open-source alternative to AWS Bedrock AgentCore, offering similar capabilities without vendor lock-in. Its MCP Firewall and model gateway provide fine-grained control over agent behavior, while its focus on human-in-the-loop workflows ensures that critical actions are always reviewed. The project is free to use and self-hostable, making it accessible to startups and enterprises alike.
- Category
- AI Agents
- Pricing
- Free
Problem solved
Preloop solves the challenge of managing and securing multiple AI agents across an organization by providing a centralized control plane. It routes agent tool calls through a firewall, governs model traffic with budgets and attribution, and offers human approvals, observability, and audit trails, ensuring compliance and reducing risk.
Use cases
- Govern tool access for AI coding agents like Codex CLI and Cursor
- Route LLM traffic through a central gateway with per-team budgets
- Approve sensitive agent actions from Slack or mobile devices
- Maintain searchable audit trails for AI agent activity
- Enforce regulatory compliance (EU AI Act, NIS2, DORA) for AI agents
- Onboard new MCP-compatible agents with a single command
- Monitor runtime behavior of AI agents across a fleet
Audience
- Developers
- ML Engineers
- Data Scientists
- Enterprise
- Startups
Pros
- Open source and free to use
- Self-hostable for full data control
- Supports a wide range of AI agents via MCP
- Built-in MCP firewall and model gateway
- Human approvals and audit trails for compliance
Cons
- Requires self-hosting infrastructure and maintenance
- Setup complexity for large agent fleets
- Dependent on the maturity of the MCP ecosystem
- May have limited documentation initially
Official website: https://preloop.ai/
Canonical LambdaBase page: https://www.lambdabase.com/tools/preloop