Skip to content
Menu

Open source, MIT licence

AI visibility, open-sourced

Track how your brand appears across ChatGPT, Gemini, Claude, Perplexity and Grok, and check whether what they say is correct.

We've taken the AI Visibility module of the Agentcy platform and made it free for anyone to download, run, inspect and build on. There's no proprietary score. Every number comes from a method you can reproduce and show a client.

Announced at the AI for PR Conference in London, June 2026.

For PR teams, being mentioned is no longer enough. What an AI assistant says about you matters as much as whether it names you at all. A brand can turn up in answers constantly and still be misrepresented, or be cited as a source without ever being named in the response.

What the platform measures

Run prompts against several models and see how your brand and your competitors show up in the answers.

Visibility
How often a brand is mentioned or recommended across the leading AI models.
Accuracy
Whether those mentions are factually correct and well positioned, checked claim by claim against a brand source of truth.
Citations
Which sources the models cite, and whether those citations turn into brand mentions.
Sentiment
Whether mentions are positive, neutral or negative in tone.
Competitor benchmarking
How rivals are positioned in the same answers about your category.
Trends over time
How visibility, accuracy and positioning shift across scheduled, repeatable runs.

The accuracy layer

Visibility tracking, citation analysis, sentiment, competitor benchmarking and trend monitoring are table stakes now. What's been missing is a way to check whether what AI says about your brand is factually correct.

The platform breaks each brand description into separate, testable claims, checks them against approved brand material and flags every divergence: wrong facts, outdated positioning, missing context and competitive misattribution. Sentiment is not accuracy, and a cited domain doesn't prove the answer is right.

Every score can be traced back to its evidence, because you're running it yourself:

  • The prompt that was asked
  • The answer the model returned
  • The exact URLs the model cited
  • A timestamp for every run
  • Each claim, checked against your source of truth

Why we open-sourced it

The PR community needs to own AI visibility as a communications channel for reaching its audience. AI answers are becoming where reputations are made and lost, and that infrastructure shouldn't sit behind a paywall or a score nobody can inspect.

The market is already pushing back. A Digiday investigation (1 May 2026) found marketers questioning expensive AI visibility tools after inconsistent results and a limited ability to prevent the hallucinations and misattributions they were bought to address.

By making the measurement layer open source, we're giving agencies a foundation they can inspect, verify and run themselves. It was built with input from senior practitioners across PR, corporate communications and public affairs. If PR is going to be taken seriously as a performance discipline, accuracy has to be part of what we measure.

Much of the hype about GEO centres on visibility, but for the work we do with clients in the corporate, government and not-for-profit space, being portrayed accurately in AI answers is even more important. Tools that just focus on visibility can't give corporate communications professionals the data they need to protect and improve reputations.

Stuart Bruce, The PR Futurist, Co-CEO, Purposeful Relations

Agencies are paying for tools that measure whether a brand appears in a result, not whether what it says is correct. That's the gap we built this to address. Open-sourcing the measurement layer means any agency can verify the results themselves, without trusting a black box they didn't build and can't inspect.

Tom Fry, CTO, Agentcy

Run it on your own machine

The whole app runs locally on SQLite, with no Docker and no cloud database, and the only cost is the AI tokens you use. When you're ready to give clients access, switch to PostgreSQL and deploy it on your own infrastructure.

  • Bring your own keys. Connect the providers you choose: OpenAI, Google, Anthropic, Perplexity or xAI.
  • Host it yourself. Use SQLite for free local testing and PostgreSQL in production, deployed with Docker on your own servers.
  • Build on it. It's MIT-licensed, and the repository has full documentation and setup instructions.

Read the setup guide on GitHub

You need Node.js 18 or later, Python 3.10 or later (for the background worker) and at least one AI provider API key.

Start where you already work

How do you want to work?