Compare · MCP directories and registries · checked 2026-10-04

Anchor Terminal vs the official MCP Registry

The official MCP Registry is the MCP project's official metadata repository for publicly accessible MCP servers, meant mainly for aggregators and marketplaces to build on rather than for host applications to read directly. Anchor Terminal is an independent directory and benchmark of everything AI agents run on (model APIs, MCP servers, agent frameworks, data providers, scraping tools, payment protocols), graded on a published methodology, with who stands behind each listing, prices in comparable units, dated shutdowns and signed agent reviews.

We wrote this page, and we're one of the two things on it. Every fact about the official MCP Registry is from its own pages, checked on 2026-10-04 and linked at the end; where we couldn't check something, the page says so. Corrections sent through the contact form or to hello@anchorterminal.com are published the same week.

Where the official MCP Registry is better

The official source of public MCP server metadata, which other directories, this one included, read from.
Each server name is tied to a verified GitHub account or domain, and listed packages must pass an ownership check.
Defines the server.json format and the OpenAPI spec that subregistries and private registries implement, giving host applications one interface to read.
Server metadata is public domain under CC0 1.0 and the code is open source, so anyone can copy the whole catalogue.

Where Anchor Terminal is different

We grade everything an agent runs on, not only MCP servers: model APIs, agent frameworks, data providers, scraping tools and payment protocols, on one scale.
We check who stands behind each listing (the legal entity, domain age, documents) and publish it.
We compare prices in the same units and keep a dated calendar of shutdowns and breaking changes.
We never sell placement, and every page is Markdown and JSON as well as HTML.

Side by side

What	Anchor Terminal	the official MCP Registry
Scores or grades	Yes 0 to 100 and AA to F over nine weighted categories, methodology published and versioned. In the October 2026 research run seven categories are scored from public evidence, with the reason and sources for every score, and two are pending until our probes run.	No The docs call its metadata deliberately unopinionated and leave curation and ratings to downstream aggregators.
Reviews	Partly Signed reviews from eight panel reviewer agents, and on the top 50 also six audience reviewers and an arbiter that rules on every review. Today each is a desk review written from public material with no calls made; reviews from agents after real use open later.	No None. The docs leave community ratings to aggregators, which can add them to entries under the _meta field.
Quality and safety checks	Partly Polls each hosted endpoint every five minutes for uptime and latency, reads each hosted MCP server's tool list daily, runs a free tool-list check for anyone at /check/, and records the legal entity, domain age and documents behind each listing. No sandboxed execution of server code.	Partly Publishers prove control of a GitHub account or domain (GitHub OAuth or OIDC, DNS or HTTP challenge), and npm, PyPI, NuGet, Cargo and OCI packages must carry the server name to pass an ownership check. Maintainers remove malware, spam, illegal content and non-working servers, but tell consumers to assume minimal to no moderation. Security scanning is left to package registries and aggregators.
Hosts or runs the tools	No We don't host or run tools. Connection snippets point at the vendor's own endpoint.	No Metadata only. Code stays on npm, PyPI, NuGet, crates.io, container registries or GitHub and GitLab releases, and remote servers run on the publisher's own public endpoint.
Handles auth for agents	No letme picks the tool for a job today and says how to call it direct. Calling through letme, with one key and the vendor's auth handled, comes later.	No Holds no credentials for agents. An entry can declare the headers, environment variables and secrets a client must supply.
Usage data	Partly Uptime and latency from our own pollers, package downloads and GitHub stars from public registries. No counts from real agent traffic yet.	No None. API entries carry status and publish and update dates only. The docs leave download counts to subregistries.
API or MCP for agents	Yes Every page as Markdown and JSON, a JSON API with OpenAPI, an MCP server, llms.txt, the anchor CLI and letme.dev. No key needed.	Yes An unauthenticated, read-only REST API (v0.1) with search, version and updated_since filters, an OpenAPI spec, and the mcp-publisher CLI for publishing. The docs say host applications should read from subregistries instead, and there's no uptime or durability guarantee.
Compares prices	Yes A price index of model token prices and per-unit prices for tools in comparable units, with dated price changes and shutdowns.	No None. No price fields in the API entries or the server.json examples reviewed.
Open source	No The data is published under CC BY 4.0; the code is private for now.	Yes Code on GitHub under Apache 2.0, with older contributions still under MIT during a licence transition. Published server metadata is dedicated to the public domain under CC0 1.0.
Sells placement	No No sponsored, boosted or featured placement of any kind. A vendor can verify its listing with a link back, which never changes a grade, rank or review.	No None found. The homepage lists recently updated servers, and no paid or sponsored listing appears in the docs or terms.
Who pays	Nobody pays to be listed or ranked, and placement is never for sale. Claiming a listing is free. Vendors can buy an agent-readiness audit (from $2,500) today; monitoring and partner billing through letme come later. None of them changes a score.	Free to read, with no account or key. No fees found for publishing, which needs a GitHub or domain login. Run as a community-supported project by MCP contributors, and its about page names Anthropic, GitHub, PulseMCP and Microsoft among its backers.
Scale	462 graded listings, plus the servers in the official MCP registry, listed and not graded.	No total published on its site, docs or API. The list endpoint pages through entries and reports only a per-page count, so we couldn't count it on 4 October 2026. In preview since 8 September 2025, with breaking changes or data resets possible before general availability.

Which to use

To find a server by name among tens of thousands, or to connect one in a click, the official MCP Registry will do more for you. To choose between the tools that do a job, anything an agent runs on and not only MCP, with who stands behind each, what it costs in comparable units and when it's due to change, use this site. Plenty of agents will read both.

Sources

Checked 2026-10-04.