MCS Certification

MCS (MyClaude Certified Standard) is the quality system that tells buyers what they are getting before they install.

Every product on MyClaude carries an MCS level. It is not decorative — it drives search ranking, enables premium pricing, and builds the trust that makes a marketplace work. This page explains what each level means, what it requires, and why it matters.

Why MCS exists

A marketplace without quality signals becomes a junk drawer. Buyers cannot distinguish a polished, tested product from a placeholder someone published in five minutes. Creators who invest real expertise in their products get buried under volume. Prices race to zero because nothing differentiates products on quality.

MCS solves this by making quality visible. When a buyer sees an MCS-3 Gold badge, they know the product has been validated against a specific, meaningful standard. When a creator invests the work to reach MCS-3, they earn tangible benefits — better placement, justified pricing, and buyer trust.

The system is designed to be progressive. MCS-1 is achievable in your first publish. MCS-3 is aspirational — it represents genuine craft. The gap between them is intentional. It gives creators something to build toward, and gives buyers a reason to pay more.

The three levels

MCS-1 Bronze — Publishable

The minimum bar for marketplace listing. Every product on MyClaude is at least MCS-1. Validation is fully automated — the CLI and Creator Engine can verify all requirements without human review.

Requirements:

• Valid vault.yaml with all required fields (name, version, type, description, entry)
• Primary content file exists and is non-empty (SKILL.md, AGENT.md, SQUAD.md, etc.)
• README.md present with setup instructions
• No secret or credential patterns in any file
• No content policy violations
• All file references in vault.yaml resolve to actual files
• No placeholder content (TODO, lorem ipsum, FIXME markers)

What the automated checks verify:

myclaude validate
  structure-check     correct files for product type?
  metadata-check      all required fields present and valid?
  reference-check     all file references resolve?
  content-length      minimum viable content (not empty stubs)?
  secret-scan         no credentials or API keys?
  readme-check        README present with required sections?
  placeholder-scan    no TODO/lorem/placeholder content?

MCS-1 answers one question: "Is this product complete enough to not waste a buyer's time?" If all checks pass, the product is publishable.

MCS-2 Silver — Quality

Products that demonstrate craft, thoroughness, and real testing. MCS-2 validation is semi-automated — most checks run automatically, but some require creator self-reporting.

Requirements (everything in MCS-1, plus):

• At least 3 examples or exemplars covering different use cases
• Anti-patterns section documenting what NOT to do
• Tested with at least 5 different user intents or scenarios
• Error handling for edge cases
• No shallow or generic content — depth appropriate to the product's domain
• Consistent output format across different inputs
• Average rating of 4.0 or higher (once reviews exist)
• At least 10 downloads (proves real-world usage)

What the checks verify:

myclaude validate --level=2
  All MCS-1 checks
  exemplar-count      >= 3 examples?
  anti-pattern-check  anti-patterns section exists?
  depth-check         content exceeds minimum word count?
  format-check        output format is specified and consistent?
  [self-report]       tested with 5+ intents?

The rating and download thresholds are checked by the marketplace, not the CLI. A product can be locally validated at MCS-2 for content quality, but the badge only appears on the marketplace once the community thresholds are met.

MCS-2 answers a harder question: "Is this product good enough that buyers will recommend it?"

MCS-3 Gold — State-of-the-Art

The highest tier. Products that set the bar for their category. MCS-3 validation requires agent-assisted review — an AI agent analyzes the product for depth, composability, and resilience to edge cases.

Requirements (everything in MCS-2, plus):

• Deep knowledge base encoded in references (domain expertise that AI alone could not generate)
• Adaptive modes — works in different contexts, constraints, or user skill levels
• Composable — works well alongside other MyClaude products
• Stress-tested against ambiguous, adversarial, and edge-case inputs
• Progressive depth modes (surface for quick use, deep for thorough analysis)
• Verified author (identity confirmed through marketplace verification)
• Consistent update history — active maintenance, not a single publish and abandon
• Community engagement — responds to reviews, ships improvements based on feedback

What the checks verify:

myclaude validate --level=3
  All MCS-2 checks
  [agent] depth-review         references contain real expertise?
  [agent] composability-test   works with standard products?
  [agent] stress-test          handles ambiguity and adversarial input?
  [agent] progressive-depth    surface/dive/radical modes work?
  [agent] token-efficiency     reasonable context usage?

Agent review means a specialized AI reviewer analyzes your product content. It is not a rubber stamp — the agent evaluates whether your product encodes genuine domain expertise or is simply reorganized generic knowledge. Products that could be fully generated by AI without domain input will not pass MCS-3.

MCS-3 answers the most demanding question: "Is this product the best available in its niche?"

Requirements at a glance

Benefits of certification

Higher MCS levels unlock tangible benefits on the marketplace.

The search ranking boost is the most impactful benefit for most creators. When a buyer searches for "code review," MCS-3 products appear before MCS-2 products, which appear before MCS-1 products, all else being equal. This is not a minor tiebreaker — it is a primary ranking signal.

Premium pricing tracks with MCS level. The marketplace does not enforce pricing by tier, but buyer behavior does. MCS-3 products routinely command 5-10x the price of MCS-1 products in the same category because buyers have a concrete quality signal justifying the price.

How to achieve each level

Reaching MCS-1

Every creator starts here. If you follow Publishing Your First Product, you will naturally produce an MCS-1 product.

Checklist:

1. Run myclaude init to scaffold your project with the correct file structure
2. Fill in vault.yaml with accurate metadata — no placeholder descriptions
3. Write your primary content file (SKILL.md, AGENT.md, etc.) with real, complete instructions
4. Write a README.md with: what the product does, how to install, how to use, a basic example
5. Run myclaude validate and fix any errors
6. Run myclaude publish

Common reasons MCS-1 validation fails: placeholder content left in scaffolded files, missing README, description exceeds 160 characters, unquoted version string.

Reaching MCS-2

MCS-2 requires investment beyond the minimum. Budget 2-4 hours of refinement on top of your initial publish.

Checklist:

1. Add at least 3 examples covering different use cases — not variations of the same case, but genuinely different scenarios
2. Add an anti-patterns section: what should your product NOT be used for? What inputs produce bad results?
3. Test with 5 different user intents. Document what you tested and what happened. Fix any inconsistencies
4. Add error handling: what does your product do with ambiguous input? Incomplete context? Edge cases?
5. Review the output format: is it consistent across all your test cases? Tighten the format specification if not
6. Run myclaude validate --level=2 and fix any gaps
7. Publish the update. Wait for 10 downloads and a 4.0+ average rating

The download and rating thresholds will come naturally if your product is useful. Focus on quality — the metrics follow.

Reaching MCS-3

MCS-3 is not a weekend project. It represents the kind of expertise that takes months or years to develop, packaged into a product that delivers that expertise reliably.

Checklist:

1. Build a genuine knowledge base. Encode domain expertise that cannot be reproduced by asking an AI to "write a skill about X." If your references section reads like a textbook summary, it is not deep enough
2. Implement adaptive modes: surface mode for quick answers, deep mode for thorough analysis, and optionally a radical mode for unconventional approaches
3. Test composability: install your product alongside other MyClaude products and verify they work together without conflicts
4. Stress-test: feed your product ambiguous input, contradictory instructions, adversarial prompts. Document how it handles each
5. Verify your author account through the marketplace verification process
6. Ship at least 2-3 updates based on real user feedback before pursuing MCS-3
7. Run myclaude validate --level=3 which triggers agent review

The agent review is the gate that separates MCS-2 from MCS-3. It evaluates whether your product encodes genuine expertise, handles stress gracefully, and works as part of the broader ecosystem. Expect to iterate.

What disqualifies a product

Certain conditions prevent MCS certification or can cause an existing certification to be downgraded.

Blocks certification entirely:

• Content policy violations (hate speech, illegal content, deceptive claims)
• Secret or credential exposure in product files
• Plagiarism of another creator's product
• Fraudulent reviews or artificial download inflation

Downgrades existing certification:

• Abandonment: no updates or review responses for 6+ months drops MCS-3 to MCS-2
• Rating degradation: average rating falling below 4.0 drops MCS-2 products to MCS-1
• Policy violations discovered post-publish: immediate downgrade to MCS-1 pending review

Will not pass MCS-2+ validation:

• Generic content that any AI could generate without domain input (the Anti-Commodity Gate)
• Shallow examples that are variations of the same scenario
• Missing error handling
• Products that try to do everything instead of one thing well

The Anti-Commodity Gate deserves emphasis. Before publishing any product at MCS-2 or above, the validation system checks: does this product encode expertise that differentiates it from what a buyer could get by simply asking Claude? If the answer is no, the product stays at MCS-1 until the creator adds genuine depth. This is coaching, not rejection — the validation provides specific feedback on what depth is missing.

MCS for different product types

Each product category has type-specific MCS requirements. The tables below show what each level requires for the most common types.

Skills

Agents

Squads

Workflows

See Product Categories for the complete list of category-specific requirements.

Pricing and MCS

The marketplace does not set prices — creators do. But MCS level shapes buyer expectations.

These are patterns from marketplace behavior, not rules. A free MCS-3 product is fine. A $49 MCS-1 product will struggle because buyers expect MCS-1 products to be lower-priced.

The general principle: MCS level justifies price. If you want to charge premium, invest in quality first. See Monetization Guide for detailed pricing strategies.

Related pages

• CONDUIT Publishing Protocol — how MCS flows through the publish pipeline
• vault.yaml Specification — the mcs_level field in your manifest
• Publishing Your First Product — achieve MCS-1 with your first publish
• Monetization Guide — pricing strategies by MCS tier
• Product Categories — category-specific MCS requirements