About MCP100

Independent editorial recommendations for AI tools and Model Context Protocol (MCP) servers.

Last updated: August 9, 2026

MCP100 helps people discover AI tools and Model Context Protocol (MCP) servers they can confidently use in real work.

Instead of asking "Which MCP is most popular?", we ask a more useful question:

Which MCP is the right choice for this job?

Every recommendation is based on editorial evaluation—not sponsorships, paid placement, or popularity alone.

Why MCP100 exists

The AI tooling ecosystem is evolving faster than most people can keep up with.

New MCP servers appear every week. Existing projects change maintainers, add capabilities, become inactive, or disappear altogether.

Directories make discovery easier, but discovery is only the beginning.

People still need answers to questions like:

  • Is this actively maintained?
  • Can I trust the publisher?
  • What permissions does it require?
  • How much evidence supports using it?
  • Is it suitable for production?
  • What information is still unknown?

MCP100 exists to help answer those questions through independent editorial recommendations.

What a recommendation means

Being listed does not automatically mean being recommended.

A Selected MCP has enough attributable evidence for MCP100 to recommend it for a defined use case today.

A Watchlist MCP shows potential, but important questions remain unanswered. It may become a recommendation as new evidence becomes available—or it may not.

A Removed MCP is retained as dated history after it leaves the public recommendation set; its reason and supporting sources remain visible instead of being silently erased.

Neither label guarantees security, compatibility, future maintenance, or suitability for every environment.

Recommendations are editorial judgments, not certifications.

MCP100 reviews publisher documentation and attributable evidence. It does not execute third-party MCP packages during documentation checks.

How recommendations are made

Every MCP follows the same editorial process.

Admission

Does this project meet the editorial baseline required to be considered?

Coverage

Does it improve the collection by representing a meaningful workflow or capability?

Ranking

Once editorial decisions have been made, ranking determines presentation order—not recommendation status.

Popularity may influence presentation, but it never determines whether an MCP is recommended.

What we evaluate

Recommendations are supported by attributable evidence rather than opinion.

Editorial evaluation considers factors such as:

  • Practical task fit
  • Publisher identity and provenance
  • Requested permissions
  • Security-related evidence
  • Maintenance history
  • Installation experience
  • Supported clients
  • Documentation quality
  • Community adoption

No single metric determines a recommendation.

Editorial principles

Everything published on MCP100 follows a small number of consistent principles.

Evidence before scores

Evidence matters more than popularity metrics.

Fit before popularity

The best tool for a workflow is not always the one with the most stars.

Unknown means unknown

Missing evidence is never treated as proof.

Recommendations evolve

Recommendations change as projects evolve and new evidence becomes available.

Commercial independence

Payment cannot buy a recommendation.

Advertising, sponsorships, partnerships, affiliate relationships, or commercial agreements never determine whether an MCP is recommended.

Editorial decisions remain independent from commercial relationships.

MCP100 is an independent project and is not affiliated with Anthropic or any MCP publisher.

Transparency

Recommendations should explain themselves.

Where possible, MCP100 keeps supporting evidence, editorial reasoning, maintenance signals, and setup guidance visible so readers can understand why a recommendation exists—not simply that it does.

Frequently asked questions

Why aren't there exactly 100 recommended MCPs?

Because recommendations are evidence-based rather than quota-based. If only the projects that meet the editorial standard are recommended, that's enough.

Can companies pay to be recommended?

No. Commercial relationships never influence editorial decisions.

Why isn't a particular MCP included?

Some MCPs are still under review, some lack sufficient evidence, and others may not currently satisfy the editorial standard.

How often are recommendations updated?

Recommendations are reviewed as projects evolve and new evidence becomes available.

Final thought

MCP100 isn't trying to list every MCP.

It's building a collection you can trust.