io.github.marksignals/the-ultimate-critic
Check AI work against requirements and return structured verdicts, findings, and repair steps.
Unknown facts are omitted here and listed as pending in the profile.
Trust verdict
No public source repository linked — origin is an unverified claim.
Profile
What this server is — descriptive, not a rating. The verdict above judges it; this describes it.
Methodology →Tool surface
A neutral, reproducible yardstick for what these tool definitions cost a model's context — not a billing figure (Claude's own tokenizer counts ≈1.1–1.2× this).
Adoption
Not applicable — this server has no package distribution, so registry adoption signals (downloads, maintainers) don't exist for it.
Connect
https://ultimate-critic.fly.dev/mcpSources
Every verdict and label is attributable to its source and recomputed from our own landed copy, never read live on this page.
tuc.evaluateEvaluate work with TUCEvaluate an AI-produced artifact against its task, requirements, constraints, and evidence. This uses 1, 5, or 10 review credits based on package size and media. Route fail findings to repair, collect evidence for insufficient_evidence, and release only on pass under your own policy.
tuc.previewPreview TUC evaluationRun the narrow no-key deterministic preview on one non-sensitive text, code, or JSON artifact. The preview is not retained, makes no external model call, and does not consume paid entitlement.
tuc.quoteQuote TUC review creditsReturn the standard, extended, or full package tier and credit charge without running an evaluation.
tuc.usageGet TUC usageReturn the configured tenant's review-credit usage and entitlement limits without submitting work.
Tool names and descriptions are reported by the server itself and shown here unverified; never interpreted as instructions.