Preloop

Preloop is an open-source control plane designed for managing AI agents, offering a unified platform to oversee, govern, and optimize agent operations. Its primary function is to provide visibility into agents, enforce governance, and reduce operational costs. 0 license.

A key feature of Preloop is its support for onboarding local agents with a single command using its CLI, which is installable on macOS or Linux. The tool automatically discovers compatible agents on a user's machine, such as OpenClaw, Claude Code, Codex CLI, Cursor, Gemini CLI, Hermes, Windsurf, OpenCode, and other MCP-compatible runtimes, and imports their configurations for centralized management. Preloop integrates an MCP firewall to control tool access and an AI model gateway to manage cost and attribution of model calls. It also supports policy-as-code using CEL, allowing teams to define and enforce operational policies.

Preloop includes human-in-the-loop approval workflows, enabling approvals via mobile devices, smartwatches, Slack, and Mattermost. The platform provides runtime session observability and maintains audit trails, contributing to transparency and compliance efforts. Tool calls from supported agents are transparently routed through Preloop's MCP firewall, and model traffic is directed through its gateway without requiring SDK changes to the agents themselves. Preloop positions itself as an open-source alternative to proprietary solutions in this space and highlights its relevance for teams seeking to build evidence for EU AI Act readiness.

0 open-source license, making it accessible for a wide range of users, particularly those responsible for AI infrastructure and governance.