56% of the MCP servers we reviewed are run by the vendor

71 of 127 reviewed entries come from the company whose product they expose. Official status changes what you should expect — and what you should not.

2026-11-03 2 min read The avbbc desk mcpdatamaintenance

When we review a server we record who runs it. Of 127 reviewed entries, 71 are published by the vendor whose product they connect to — Postman, Pulumi, Cloudflare, Rollbar, Doppler, Vanta, and so on.

That number is higher than most people assume, and it is worth understanding what it does and does not buy you.

What official status buys

**The install path matches the product.** A vendor server knows its own authentication scheme, its own rate limits, and its own API quirks. It is written by someone with access to the internals.

**It ships with the product.** Official servers tend to be documented where the product is documented, which means an announcement when the API changes. Community servers find out when something breaks.

**There is a support relationship.** Not always good support, but a company rather than a person. That matters when the alternative is a maintainer who stopped replying.

What it does not buy

**Safety.** An official server has whatever permissions it asks for. Pulumi's server can preview infrastructure changes; it can also be given a token that applies them. The vendor made a connector, not a guardrail.

**Restraint.** Official servers tend to expose everything the API offers, because the vendor's incentive is completeness. A community server written for one workflow is often narrower, which is sometimes better: fewer tools means better tool selection.

**Longevity.** Companies deprecate products. When they do, the connector usually goes with it, and the shutdown notice goes out on a channel you do not read.

How to read the badge

In our directory, official servers are labelled. Our advice is to treat the label as a prior, not a conclusion: it predicts documentation quality and install-path correctness, which is most of the friction. It says nothing about what the server can destroy if an agent misreads a request.

The remaining 56 reviewed entries are community projects. Some are excellent and better documented than vendor equivalents — but they carry a maintenance question the vendor ones do not, and the description on each entry says so.

The pattern worth watching

Official share is rising. As it does, the community tier will be pushed toward niches vendors do not care about: internal tools, small SaaS, and glue. That is where the interesting servers will live and where reviewing them will matter most.

Keep reading

All notes