SilentCritique.
Evidence-backed AI trust

Run a verdict
on your work.

We hold public AI to a hard, evidence-first standard. Start a scoped evaluation for your product, agent, or generated work, then publish the score only when the evidence is ready.

Run a VerdictBrowse 15 real verdicts

15 public verdicts · every score backed by a reproducible test.

Live verdict

88

MCP server

MCP Time Reference Server

Narrow, deterministic utility

The Time reference server passed a real MCP smoke test for current-time lookup and timezone conversion. It scores well because the task is narrow, deterministic, and low-side-effect, though invalid input and localization behavior were not deeply tested.

See the evidence

How a verdict works

01

Scope it

Name the claim, target, and evidence surface. Public verdicts start with a testable question.

02

We test, not guess

Every claim is tied to real evidence — what was checked, what passed, what failed.

03

You get a verdict

A clear score with strengths, failure modes, and the one next move that matters.

Real verdicts, real evidence

These are live smoke tests we ran against public AI tools — not opinions. Open any one to see the exact checks behind the score.

Browse all 15

MCP server

87

Calculator MCP Server

The Calculator MCP server passed a real smoke test: it evaluated arithmetic correctly and, importantly, rejected a code-execution attempt disguised as an expression with an AST-level unsupported-operation error. It scores high because the task is narrow, deterministic, and the observed input handling refused non-arithmetic code.

See the evidence

MCP server

80

Chroma MCP Server

The Chroma MCP server passed a real smoke test: it created a collection, embedded and stored a document, retrieved it by semantic query, and failed cleanly when asked about a collection that does not exist. It scores well as a vendor-maintained memory primitive, tested here only in ephemeral in-process mode.

See the evidence

MCP server

77

Excel MCP Server

The Excel MCP server passed a real smoke test: it created a workbook, wrote a data range, and read the same cells back with full per-cell metadata, all without Excel installed. The score is solid for the verified round-trip, with caution because only 3 of 25 exposed tools were exercised and file paths are not sandboxed by the server.

See the evidence

How we score

The rubric, evidence rules, and what a score does and does not mean.

For agent builders

Crawlable protocol, OpenAPI, and llms.txt so agents understand the contract.

Pricing

Free demo practice. $4.99 per instant critique. Every charge logged and visible.

Hold your own work to the standard.

Start a public verdict when you need an evidence-backed trust artifact. Use Instant Critique when you need a private report now.

Run a VerdictTry Demo Ritual
Run a VerdictBrowse 15 verdicts