Owned by Nous Research or the service owner.
Trust is an explainable processWhat we know, what we interpret,
What we know, what we interpret,
and what we do not claim.
Our goal is to show the origin and limits of each claim before your agent can reach a file, account, or external service.
Content methodology
Five stages for every decision.
- 1
Collect from the source
Official documentation, Official MCP Registry, Hermes Registry, and original project repositories.
- 2
Separate fact from interpretation
Names, descriptions, publishers, and versions remain source data. Plain-language copy is clearly editorial.
- 3
Never invent a command
If the registry does not support a safe Hermes setup path, we send the reader to the publisher.
- 4
Explain risk without guaranteeing safety
Transport, authentication, and write access are decision signals, not a security audit that never happened.
- 5
Keep correction open
Every resource links upstream, and signed-in users can send a correction to the admin dashboard.
Source tiers
The label matters.
We do not confuse official, registered, community, and editorially reviewed.
Published in an official registry with publisher and version data, but not owned by Hermes.
An independent project linked to a known source.
A selection or explanation from our platform, with source and review date.
What did we learn from the market?
Take the useful pattern, then apply it transparently.
Hermes Atlas
Pattern: Ecosystem map, curated lists, use cases, handbook, and reports.
Our implementation: A broader directory, collections, solutions, comparisons, and source-specific pages.
Glama
Pattern: Clarity around source, tools, permissions, and verification methodology.
Our implementation: Source cards, risk explanation without invented scores, and a safe first test.
Smithery
Pattern: Capability search, toolbox, clear setup, and a publisher workflow.
Our implementation: Capability search, stacks, sourced commands, and a submission review flow.
MCP.so
Pattern: Servers, clients, skills, agents, and workflows in one discovery surface.
Our implementation: A unified typed catalog connected to academy and projects.
PulseMCP
Pattern: Use cases, editorial content, updates, and community submission.
Our implementation: Solution center, Arabic digest, submission, and content reporting.
SkillsMP / skills.sh
Pattern: Browse by category, creator, and adoption while preserving GitHub provenance.
Our implementation: Type, publisher, registry, and source remain visible, with goal-based collections.
Trust questions
Is every directory resource safe?
No. Presence in a public registry makes metadata discoverable; it does not prove the code or permissions fit your environment.
Is this a copy of the official documentation?
No. We write an original explanation and retain the source title, description, and link for verification.
What does last verified mean?
It is the date of the registry or source snapshot, not proof that the resource ran on a device like yours.
Can a publisher buy a higher ranking?
Not in editorial results. Any future sponsorship must be clearly labeled and cannot change verification status.