Independent AI Certification Authority
Trust claims in AI
should be verifiable,
not asserted.
Fidensa independently certifies AI capabilities — MCP servers, skills, agent rules files, hooks, sub-agents, and the plugins that compose them — through behavioral testing, adversarial analysis, and supply chain verification. Every certification is a signed, portable artifact that anyone can verify against the evidence.
Eight-stage verification pipeline
Every capability passes through the same defined pipeline. Each stage produces artifacts that feed the trust score and are published in the signed contract.
Ingest
Source acquisition, build, interface extraction, provenance hash
SBOM & Supply Chain
Dependency tree mapping, vulnerability cross-reference (syft, grype, osv-scanner)
Security Scan
Static and behavioral analysis (Cisco mcp-scanner, skill-scanner)
Functional Test
Valid, edge, error, and LLM-generated test cases in isolated sandbox
Adversarial Test
Structured attack library — 6 categories, 55 patterns — with impact-based finding classification
Behavioral Fingerprint
Per-tool timing (p50/p95/p99), error rates, resource profiling, baseline for drift detection
Certify
Contract assembly, trust score from eight weighted signals, ES256 signing, certification issuance
Community Validation
Crowdsourced experience reports with recency-weighted confirmation rates, feeding maturity and correctness signals back into the trust score
Stage 8 — The feedback loop
Stages 1 through 7 are our assessment. Stage 8 is yours. After certification, AI agents and developers report their real-world experience: did the capability work as the contract promised? Reports are recency-weighted, anti-spam verified, and feed directly into the trust score's consumer confirmation signal.
As reports accumulate, certifications mature from Initial (lab results only) through Emerging, Established, and Proven — each level requiring more reports, more unique reporters, and longer monitoring windows. A 91/A certification at Initial means excellent pipeline results with no field data yet. The same score at Proven means hundreds of consumers confirmed it over six months.
Report via the API, the MCP server, or any agent with Fidensa tools installed. No authentication required.
Live certifications
50 capabilities evaluated through the full pipeline. Real data, real findings, real scores.
Use it today
Check a capability before you install it. Add a badge to your README. Wire verification into your CI pipeline.
MCP Server
Give your AI agent access to Fidensa certification data. Check trust scores, search certified alternatives, verify artifacts — all through MCP tool calls.
{
"mcpServers": {
"fidensa": {
"command": "npx",
"args": ["-y", "@fidensa/mcp-server"]
}
}
}GitHub Action
Fail CI builds when a capability loses certification or drops below a score threshold. One line in your workflow.
- uses: fidensa/check-certification@v1
with:
capabilities: playwright-mcp
min-tier: verifiedBadges
Show certification status in your README, docs, or registry listing. Updates automatically.
Attestation API
Query certification status programmatically. No API key required for basic checks.
GET https://fidensa.com/v1/attestation/fidensa-mcp-serverBeyond vulnerability scanning
Static analysis checks whether validation code exists. Behavioral testing checks whether it actually works when someone tries to bypass it.
A recent static analysis scored Anthropic's reference MCP servers at 99 out of 100, praising their path validation layers. Our behavioral pipeline installed the same server, connected over MCP, and threw adversarial payloads at every tool. Three critical vulnerabilities. Credential exposure. Score: 60.
The guard existed. It did not hold.
Every certification Fidensa issues answers three questions that scanners cannot: what does this capability actually do, what are the boundaries of its behavior, and what happens when someone tries to push it past those boundaries? The answers are published in a cryptographically signed contract that anyone can verify against the evidence.
Grounded in established standards
Every number in the scoring system traces to an established framework, a documented pairwise comparison, or a scenario-validated design choice. The full methodology is published in full.
CVSS v4.0
Finding severity ratios and deduction curves
NIST SP 800-30
Compounding risk factors and confidence modeling
SLSA v1.0
Provenance as foundational trust requirement
ISO/IEC 25010
Software quality characteristics ordering
ISO/IEC 17065
Certification body requirements and scheme structure
UL certification model
Pass/fail tiers against defined safety criteria
Certification tiers
Score and certification are separate. The trust score is an analytical summary. The tier is a judgment based on finding-severity thresholds.
Fidensa Certified
No unmitigated critical findings. No more than two unmitigated high findings. Pipeline completed.
Fidensa Verified
Pipeline completed. Findings of any severity documented. Evidence-backed contract issued.
Fidensa Evaluated
Pipeline ran with partial coverage. Incomplete data. Contract documents what was found.
What's ahead
Fidensa launched with 50 certifications across the MCP ecosystem. The pipeline is live, the methodology is published, and we're expanding.
Expanding the certification catalog
More MCP servers, skills, and plugins from across the ecosystem. If you use it, we should certify it.
Publisher certification submissions
Self-service pipeline access for publishers who want their capabilities independently certified.
OpenClaw skill certification
247K+ GitHub stars. 700+ community skills. 341 malicious uploads in the first two months. The fastest-growing agent ecosystem needs pre-installation trust verification.
Enterprise governance API
Organization-wide policy enforcement. Define which tiers, scores, and risk tags are acceptable across your agent fleet.
Behavioral contract as open specification
The contract format that documents what a capability does and won't do, published as an open spec for the ecosystem.
Get started
Whether you build AI capabilities, integrate them, or govern their use — Fidensa provides the evidence layer you need.
Check before you install
Search the certification catalog or query the API to check any capability's trust status before adding it to your stack.
Browse certifications →Get certified
Certification submissions opening soon. If you publish an MCP server, skill, hook, or plugin and want early access to the pipeline, reach out.
Request early access →Trust-based workflow
Add the MCP server to your agent, the GitHub Action to your CI, or badges to your README.
Integration tools →