Preloop vs MintMCP: Open-source MCP gateway with built-in approvals

Preloop is the open-source AI agent control plane: an MCP firewall, AI model gateway, human approvals, runtime observability, and audit trail in one self-hostable platform. MintMCP is a commercial MCP gateway and AI agent infrastructure product: role-based MCP bundles, agent identities, monitoring, an LLM gateway, and inline screening, sold with per-user licensing and self-hosted options. The honest one-liner: if you want a vendor-run MCP gateway with a curated catalog and SOC 2 paperwork, MintMCP is a credible choice; if you want an open-source control plane with per-call human approvals, policy-as-code, and audit evidence, choose Preloop.

At a glance

What MintMCP does well

MintMCP has grown into a broad hosted platform:

• Fast time to value. Admins approve servers, group them into role-based bundles, and users add a bundle to their client. A customer on its homepage calls the virtual bundles "our number one feature" (mintmcp.com).
• Enterprise paperwork. "SOC 2 Type II audited and compliant with HIPAA standards", with SAML SSO and EU and US data residency (pricing).
• Agent identities. Each agent gets delegated, least-privilege credentials that can be revoked on their own (mintmcp.com).
• Monitoring and screening. Spend and tokens per team, audit export to a SIEM or OTLP, Mint Guard inline, and a private tunnel to internal servers (mintmcp.com).

If you want a vendor-operated MCP gateway with a curated catalog and compliance documentation, MintMCP is a reasonable option.

Where Preloop is different

• Open source and self-hostable. Preloop's core is Apache 2.0. Run it on your own infrastructure, read and audit the code, and adapt it. There is no requirement to send MCP traffic to a third party.
• A human can decide a single call. When a rule says a person decides, the agent waits and the reviewer gets the tool, arguments, agent reasoning, and matched policy on a phone, watch, Slack, Mattermost, email, or webhook. Async mode lets long reviews finish without breaking the agent's transport.
• Native agent actions. Native action gates for Claude Code, Codex CLI, Cursor, Copilot CLI, OpenCode, and others put shell commands and file edits on the same approval path as MCP calls. A gateway only sees what crosses it.
• Policy-as-code with CEL. Ordered YAML rules with CEL conditions on tool arguments version-control like the rest of your infrastructure.
• Runtime session observability. One timeline per agent session: tool calls, model calls, matched policies, approvers, spend, and outcomes.
• Onboarding. preloop agents discover finds local agent configurations; preloop agents onboard routes supported agents through Preloop, with backups and a dry-run preview.

When to choose MintMCP

• You want a hosted MCP gateway and catalog without operating infrastructure, with a self-hosted option negotiated in the contract.
• SOC 2 Type II and HIPAA documentation from the vendor are hard requirements.
• Role-based tool bundles are the main control you need, and per-call human approval is not.

When to choose Preloop

• Platform and DevEx teams rolling out Claude Code, Codex CLI, Cursor, Gemini CLI, OpenClaw, or OpenCode who need policy and approvals on what those agents do.
• Security teams that require a self-hostable control plane and searchable audit evidence inside their own network.
• Engineering teams standardizing on one control plane for MCP governance, model traffic, approvals, and audit.
• Open-source-first organizations that want a control plane they can read, fork, and extend.
• Teams building EU AI Act readiness evidence who need durable records of matched policies, approvers, and outcomes.

Adoption notes

If you already use MintMCP, Preloop can be introduced side by side: onboard Claude Code, Codex CLI, or Cursor on one developer machine with preloop agents onboard, and keep other MCP traffic on MintMCP. Compare the Preloop session timelines and approvals against your current data, then decide which clients to move.

FAQ

Is there an open-source alternative to MintMCP? Yes. Preloop is an open-source AI agent control plane under Apache 2.0, with an MCP firewall, policy-as-code, human approvals, an AI model gateway with budgets, runtime observability, and an audit trail.

Can I self-host my MCP gateway? Yes, with Preloop on any infrastructure. MintMCP lists self-hosted options on its pricing page.

Which approval channels does Preloop support? Mobile and watch push, Slack, Mattermost, email, webhooks, and the CLI. Microsoft Teams is not a built-in channel; a webhook can bridge to it.

How do MCP tool policies work in Preloop? Policies are YAML files with ordered rules. Each rule can match on tool name, server, parameter values, agent identity, or other context through CEL expressions, and decide allow, deny, require approval, or require justification.

How do I onboard existing agents to Preloop? Install the Preloop CLI, run preloop agents discover to find local agent configurations, then preloop agents onboard to route the supported ones through Preloop.

Try Preloop

• Start your free trial
• Book a demo
• Preloop on GitHub

Last reviewed: 2026-10-08.