About MCP Verdict
MCP Verdict is an independent intelligence platform covering the Model Context Protocol ecosystem. We profile MCP servers and clients, publish side-by-side comparisons with stated criteria and named verdicts, and maintain setup guides tested against the exact errors each client throws.
We are not a directory. Directories index what exists. We help you decide what to use — and how to use it.
Why this exists
The MCP ecosystem crossed 10,000 published servers in 2026. The registries that catalog them do not evaluate them. README files describe what a server claims to do, not how it actually performs, what it costs in context window tokens, where it fails, or whether it is still maintained. The gap between “this server exists” and “this server is worth installing” is where MCP Verdict operates.
Independence
MCP Verdict is not affiliated with Anthropic, the Agentic AI Foundation, the Linux Foundation, Microsoft, OpenAI, Google, or any MCP server or client vendor. No vendor has editorial input into our profiles, comparisons, or rankings. No one pays to move up a ranking or change a verdict.
We may earn revenue from newsletter sponsorships and, in the future, from affiliate links and lead-generation partnerships. When we do, the rule is public and non-negotiable: the link existing never changes a verdict or a rank. Monetization happens around the editorial content, not inside it. Revenue sources will be disclosed on our affiliate disclosure page the moment they go live. Our full methodology explains how editorial and commercial operations stay separate.
Who does the work
MCP Verdict is researched and maintained by Pulkit, who tracks the MCP ecosystem full-time — profiling servers, verifying claims against primary sources, testing installations across clients, and documenting the failure modes that documentation skips.
The framing matters: this is an analyst operation, not a practitioner blog. Pulkit does not build MCP servers. He evaluates them. The value is not “I made this and it works” — it is “I tested this against stated criteria, verified the facts, dated everything, and published a verdict you can check.”
Corrections are welcome and published. The goal is to be right, not to appear right.
What we cover
Server profiles. Each profile covers what the server does, how to set it up per client, real token costs, authentication requirements, failure modes, security notes, maintenance status, and a verdict. Not a README summary.
Client profiles. What each MCP client supports, where its limits are, and what you hit first.
Comparisons. Side-by-side evaluations with criteria stated before verdicts. The criteria are published. The verdict follows from the criteria. If you disagree with the criteria, you can disagree with the verdict on those grounds and tell us why.
Setup guides. Per-client installation walkthroughs tested with the exact errors each client throws, not copied from the README.
News. Ecosystem updates — server launches, deprecations, spec changes, enterprise adoption. No recaps of vendor blog posts. Original reporting on what changed, what it means, and what to do about it.
How we do it
Our methodology is published in full on the methodology page. The short version: we set criteria before evaluating, verify claims against primary sources (official documentation, GitHub repositories, specification documents), date everything, and re-check on a stated cadence. When we get something wrong, we publish a correction and update the original content.
Contact
Reach us at contact@mcpverdict.com. Corrections, tips, and server suggestions are prioritized.