Skip to content
FaultlineAlpha

We check the receipts on AI output.

Paste what an AI told you. Faultline checks its most important factual claims against live sources, scores the risk, and shows you which of those it could not verify.

faultline verify
Drop .txt or paste text5,000 remaining

Or check one of these drafts

For organisations

Faultline Enterprise

Faultline Enterprise is the full platform: a REST API with key management and multi-tenant usage, an MCP server and a Python client, deployed with your rollout, a Terraform provider and a GitHub Action, multi-model consensus verification, a tamper-evident audit log, a risk register and human sign-off, and dedicated support.

Book a 20-minute call What's included

What a check returns

One verdict per claim.

Each claim gets a verdict. The draft gets a risk band, and the weakest link is named so you know what to fix first.

1 of 5 claims contradicted

5 claims extracted · 4 checked · 1 not checked

RiskMedium
  1. SupportedThe Harbor Line ferry opened in 1998.
  2. ContradictedIt departs every 20 minutes, all day.
  3. MixedIt is the fastest crossing in the bay.
  4. UnverifiedMost commuters in the region use it.

Weakest link: “It departs every 20 minutes, all day.” Contradicted by the operator's timetable.

Illustration of the layout, not a real scan. The ferry, its claims and the score are made up.

How it works

Three steps, every check.

Faultline checks the claims and citations in AI output against real sources, across multiple AI providers.

  1. Step 1

    Extract

    Faultline pulls each factual claim out of the draft.

  2. Step 2

    Verify

    Each claim is checked against live web sources and gets a verdict.

  3. Step 3

    Risk

    The draft gets a risk band, and the weakest-link claim is named.

The four verdictsSupportedContradictedMixedUnverified

Faultline Enterprise

The full platform.

Integrate

  • REST API with key management and multi-tenant usage
  • MCP server and Python client, deployed with your rollout
  • Terraform provider
  • GitHub Action

Govern

  • Risk register
  • Tamper-evident audit log
  • Human sign-off

Verify

  • Multi-model consensus verification

Support

  • Dedicated support
Book a 20-minute call about Enterprise →

Plans

Start free. Scale to Enterprise.

  • Free

    $0

    • 5 scans/month
    • JSON & SARIF export
    • Basic verification

  • Personal

    $19/mo

    • 100 scans/month
    • Markdown export
    • All export formats
    • Email support

  • Pro

    $49/mo

    • 500 scans/month
    • PDF export
    • Batch scan
    • Team workspaces
    • Notifications
    • All providers
    • API access

  • Enterprise

    Custom

    • Scan allowance set in your contract
    • Everything in Pro
    • Multi-model consensus verification
    • Audit trails
    • Dedicated support
    Book a 20-minute call

Alpha: limits and features may change; no SLA.

Full pricing →

Faultline Enterprise Rollout: in 4 weeks we deploy Faultline Enterprise in your stack, wire it into your AI and agent pipelines, set the checks with your team, and hand over a running system with an audit trail.

Includes the Faultline Enterprise subscription. Pricing on request.

For developers

The open-source CLI and the GitHub Action.

Check a file from your terminal, gate an agent's piped output, or fail a CI build on what it finds.

Bring your own provider key. 4 providers supported:

  • Google Gemini Supported
  • Anthropic Claude Supported
  • OpenAI Supported
  • Perplexity Supported
npm install -g @nxtg/faultline

faultline scan --input report.txt

claude -p "..." \
  | faultline guard --fail-on refuted

The GitHub Action runs the same gate: uses: nxtg-ai/faultline-action@v1, and its findings upload to GitHub code scanning as SARIF.

Read the CLI docsStar on GitHub