- CollectionOversight & Approval
- Classificationagent-native
- LicenseApache-2.0
- MCPAvailable
- Agent SkillUnavailable
- InterfacesMCP · REST · CLI · API
- Verified2026-09-30
What it does
Preloop is a self-hostable (or Preloop Cloud) control plane for agents that already exist on a machine. It unifies an MCP firewall (allow / deny / require approval / require justification with YAML + CEL), an OpenAI- and Anthropic-compatible model gateway with budgets and attribution, human approvals (mobile, watch, Slack, Mattermost, email, webhook), runtime session timelines, and audit evidence.
Agent-facing interfaces
- Interaction pattern: CLI that rewrites local agents onto the control plane
- Catalog protocols: MCP, REST, CLI, API
- MCP: ✅ Available (firewall / managed MCP)
- Agent Skill: ⚠️ Not published as a standard SKILL.md package
- Onboarding type:
cli
How agents onboard
Interaction pattern: CLI that rewrites local agents onto the control plane
curl -fsSL https://preloop.ai/install/cli | sh
preloop signup
# or: preloop login --url http://localhost:3000
preloop agents discover
See the full onboarding section in the dossier.
Protocol surface
| Interface | Detail |
|---|---|
| CLI | preloop signup, login, agents discover, onboard, install-plugin, validate |
| MCP firewall | Managed MCP endpoints after rewrite |
| Model gateway | /openai/v1/…, /anthropic/v1/messages |
| Control WS | WS /api/v1/agents/control/ws |
| REST | Flow execution results, security-screen score, session APIs |
Verification
Last verified 2026-09-30.
Sources:
- https://github.com/preloop/preloop/releases/tag/v0.16.0
- https://preloop.ai
- https://docs.preloop.ai