MCP gateways ranked / Methodology

Methodology: how the MCP gateway ranking is made

The ranking rests on seven criteria, public sources and a dated review. This page explains each step so you can reproduce a score or challenge one. Last review: 11 November 2026.

Last reviewed

The ranking criteria

  1. Authentication: Does the gateway know which person and which agent made each MCP request?
  2. Credential handling: Can agents use downstream tools without ever holding long-lived secrets themselves?
  3. Policy depth: Can you allow or block individual tools, arguments and actions per user or group, not just whole servers?
  4. Audit: Is there a complete, attributable and exportable record of every tool call?
  5. Observability: Can engineers trace a session end to end, with latency, errors and usage, in the tools they already use?
  6. Deployment model: Can it run where your policies require (managed, self-hosted, local), and how much work is it to operate?
  7. Shadow usage discovery: Can it find MCP servers and AI clients people set up on their own, outside the gateway?

They follow the questions platform and security teams ask first: who is calling, where the secrets live, what an agent may do, what record is left, how to see it working, where it runs, and what is happening outside it.

Scoring

  • 2 – Strong: built in and documented.
  • 1 – Partial: present with gaps, or limited scope.
  • 0 – Not documented: not found in the public docs we reviewed.

Seven criteria at up to 2 points each give a total out of 14. Every criterion carries the same weight. "Not documented" means exactly that: the capability may exist, but a buyer reading the public docs can't confirm it, so it scores 0 until it is documented.

Which gateways are included

A product is included when it describes itself as an MCP gateway or proxy, sits between MCP clients and MCP servers, and has public documentation or source code to score it from. The current list has 8 gateways, spanning enterprise platforms, open-source proxies, an API gateway with MCP support and local developer tools.

Sources

Scores come from vendors' own public material: product pages, documentation and source repositories. Each gateway's page shows the note behind each score and links to its sources. We don't use paid placements, third-party star ratings or customer reviews, and we don't publish prices or statistics a vendor hasn't published.

Order and ties

Gateways are ordered by total score. Where totals tie, the gateway that scores higher on authentication, credential handling and policy depth (the criteria that most reduce risk when agents can act) comes first. If they still tie, a gateway built for agent and MCP traffic comes before a general gateway with MCP added, and then the project with more recent public activity in its repository or changelog comes first. Verdicts and "best for" lines then say who each option suits, because the top total isn't the right pick for every team.

Reviews and corrections

Each review is dated at the top of every page. When a vendor documents a new capability, its score changes at the next review. If you see a score that no longer matches the docs, send the source through the correction form.