Early access now open

An honesty score for
every AI answer.

Ship AI answers your team can stand behind. ClearRun shows when an answer sounds surer than its reasoning allows, and gives you a signed record when the decision matters.

Start free GitHub Docs

HONESTY SCORE

62/100

"The 2024 Stanford study found a 47% improvement in…"

⚠ Unsupported: a precise statistic and named source, with no justification given

Illustrative sample, not a live result. ClearRun measures honesty (whether confidence is justified by the reasoning), not factual correctness. Early access; independent accuracy validation in progress. Methodology.

Use it where you already work.

ClearRun scores honesty (confidence-justification alignment). It is not a fact-checker. Get an API key to call the hosted API.

MCP server

Add ClearRun to Claude Desktop, Claude Code, Cursor, VS Code, or any MCP client.

@clearrun/mcp Setup docs

{
  "mcpServers": {
    "clearrun": {
      "command": "npx",
      "args": ["-y", "@clearrun/mcp"],
      "env": {
        "CLEARRUN_API_KEY": "YOUR_CLEARRUN_API_KEY"
      }
    }
  }
}

CLI

Score an answer from your terminal.

@clearrun/cli CLI docs

npm i -g @clearrun/cli
export CLEARRUN_API_KEY=YOUR_CLEARRUN_API_KEY
clearrun verify --text "The capital of France is Paris."

SDK

Call the hosted API from your own code.

@clearrun/sdk SDK docs

import { createKernelClient } from '@clearrun/sdk';

const client = createKernelClient({
  hosted: { apiKey: process.env.CLEARRUN_API_KEY },
});

const scored = await client.score({
  responseText: 'The Eiffel Tower is in Berlin.',
});

Hosted API

Score, evaluate, chat, and look up signed verdicts over HTTPS. Open the API reference for routes and examples.

api.clearrun.net/docs API keys Pricing

Verify a signed record

Anyone can check a verdict id. No API key required. Confirms the stored record; does not fact-check the answer.

GET /api/verify/{id}

curl https://api.clearrun.net/api/verify/VERDICT_ID

Docs and source

Read the API docs, then open the public packages and examples on GitHub.

API docs github.com/Adepoco/ClearRun Developer guide

Version 0.2.0 adds clearrun login, so you can sign in once instead of setting a key.

Use it as a checkpoint between AI steps.

Put ClearRun between the steps of your own pipeline or agent. Each answer gets an honesty score before it moves to the next stage, and your code sets the rule: pass it on, send it back for another try, or route it to a person when the score is below the level you choose. ClearRun never blocks anything itself. You set the threshold.

import { createKernelClient } from '@clearrun/sdk';

const client = createKernelClient({
  hosted: { apiKey: process.env.CLEARRUN_API_KEY },
});
const { verdict } = await client.score({ responseText: priorAnswer });
if (verdict.score == null || verdict.score < 70) {
  // retry or flag for a person
} else {
  // continue to the next stage
}

Three things, every answer.

Score

Every AI response gets a 0 to 100 honesty score, inline, in real time.

Flags

When confidence outruns justification, ClearRun names it: unsupported claim, overconfidence, false precision, internal inconsistency, and more.

Audit

On demand, a signed, reproducible record of the honesty verdict. Verify it anytime at api.clearrun.net/api/verify/{id}.

A layered governance model runs on every answer. It scores justification quality, then the hosted kernel signs the verdict, so a ClearRun score is one anyone can verify but nobody can forge. It does not fact-check and it never blocks your output.

See it work.

Two minutes: a question goes in, the score comes back, the flag opens, the audit prints.

Demo video coming within days. Join the list below and you will get it first.

Get early access.

Questions about ClearRun

Is ClearRun a fact-checker?

No. ClearRun does not check claims against the world. It checks whether the confidence an answer projects is earned by the reasoning and evidence it gives. A well-argued answer can still be wrong, so pair ClearRun with your own source checks when truth matters.

What do I get back?

A 0 to 100 honesty score, named flags such as unsupported claim, overconfidence, false precision or internal inconsistency, and, on request, a signed audit record anyone can verify.

Will it block my output?

No. It scores and flags; you decide what to do.

How accurate is the score?

Accuracy validation is pre-registered and in progress. Until results are published, treat the score as a signal of confidence versus reasoning, not a validated accuracy metric.

Which languages?

English today.

Who makes ClearRun?

Adepoco, a California corporation.

Built over the past year by a solo founder who got tired of confident AI answers that were quietly wrong. The industry is now walking into exactly that gap.