Get This Tool
GhostUser
Summary
You built the thing, so you can't see the thing — and by the time real users tell you the onboarding form is confusing, you've already shipped three more broken flows on top of it. GhostUser runs five AI personas through your product before anyone real touches it, surfacing the friction you stopped noticing weeks ago.
Each persona — a cautious newcomer, a skeptical evaluator, a power user, a time-pressured visitor, a motivated buyer — navigates your app autonomously, flags where it gave up, and logs why. Console errors, failed network requests, and 5xx responses get caught in the same pass, so you get UX feedback and QA signal in one run. It connects directly to localhost, which means you catch issues before they leave your machine. The tool runs on your Claude API key, so cost scales with usage rather than with a seat count. Where it breaks: the feedback reflects what five hardcoded personas notice, not the distribution of your actual users.
Bottom line: Pick this when you need to pressure-test an onboarding flow before scheduling real user sessions — skip it when your research question depends on demographic or behavioral variance your actual user base carries.
Pricing Plans
Usage-BasedLast verified 2 days ago- Price
- Free (pay-per-token for Claude API calls)
- Free Tier
- Open source. You bring your own Anthropic API key. Costs cents per test.
Pay-as-you-go
Open source tool. Users pay only for API usage via their own Anthropic API key.
- Single screenshot test: ~$0.02
- Full agent run (5u201315 steps): ~$0.10 u2013 $0.25
- Haiku ~5u00d7 cheaper
- Opus ~5u00d7 more expensive
- Real cost shown after every test
View full pricing on github.com →
Pricing may have changed since last verified. Check the official site for current plans.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Community Benchmarks Community
Sign in to submit a benchmarkNo community benchmarks yet. Be the first to share a real-world data point.
Pros
Sign in to edit- Runs five behaviorally distinct personas in a single pass, so you get failure narratives that map to real user archetypes rather than a generic click trace — without scheduling a single session.
- Catches console errors, failed network requests, and 5xx responses during the same UX walk, so a broken form submission and the confusion it causes appear in the same report instead of two separate QA cycles.
- Connects directly to localhost, so you can validate a flow before it leaves your development environment — catching the broken link that would otherwise survive until staging.
- MIT-licensed and self-hosted, so your product data stays on your infrastructure and there is no vendor seat cost — you pay only for the Claude API calls the agents make.
- Provider cost scales with actual usage rather than a fixed subscription, so a solo founder running ten flows a week pays nothing when the product is quiet.
Cons
Sign in to edit- The five personas are fixed archetypes defined in the codebase. Teams building for users with accessibility requirements, non-English interfaces, or enterprise procurement behavior will find none of the built-in personas exercise those paths — and adding custom personas requires diving into the source, not a config file.
- With six commits in the repository, the project carries early-stage fragility: browser edge cases, authentication flows with MFA, and SPAs with complex client-side routing are the first places the agent loses its footing. Teams hitting those walls have no support channel beyond filing a GitHub issue.
- The feedback quality ceiling is bounded by what Claude can infer from the DOM and network log. When a team needs statistically representative signal — sample sizes, task completion rates, time-on-task across demographic segments — this tool produces five data points, not a study. At that point teams move to a proper user research platform or moderated testing, and GhostUser stays useful only as a pre-screening pass.
Community Reviews
Sign in to write a reviewNo reviews yet. Be the first to share your experience.
About
- Platforms
- Web (Figma, Claude Desktop, Cursor, Windsurf, VSCode-compatible editors), CLI/Chromium
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- 2026-06-09T11:53:31.692Z
Best For
Who it's for
- Early-stage startups and small teams without user research budgets
- Designers working in Figma who want rapid UX feedback
- Developers building web applications who want automated QA-like testing
- Product teams iterating quickly and needing pre-validation before real user sessions
What it does well
- Testing product designs before real user testing to identify friction
- End-to-end UX validation on localhost dev servers
- Catching broken links, form errors, and console issues automatically
- Validating onboarding flows and key user journeys
- Supplementing expensive human user testing with rapid AI feedback
Integrations
Discussion Community
Sign in to commentNo discussion yet. Sign in to start the conversation.
Compare GhostUser
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeCommunity Notes & Tips Community
Sign in to contributeBe the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.
Frequently Asked Questions
- Is GhostUser free?
- Yes — GhostUser is fully free to use. There is no paid tier.
- Is GhostUser open source?
- Yes. GhostUser is open source.
- Does GhostUser have an API?
- Yes. GhostUser exposes a developer API. See the official documentation at https://github.com/savkevip/ghostuser for details.
- Can I self-host GhostUser?
- Yes. GhostUser supports self-hosting on your own infrastructure.
- What platforms does GhostUser support?
- GhostUser is available on: Web (Figma, Claude Desktop, Cursor, Windsurf, VSCode-compatible editors), CLI/Chromium.
Hours Saved & ROI Stories Community
Sign in to contributeBe the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."
Curated lists that include this category
Most UX problems get found too late — after the sprint, after the launch, after the support ticket. GhostUser is an open-source, self-hosted agent that autonomously navigates your web application as five distinct AI personas, each with a defined mindset and goal. The core workflow: point it at a URL (including localhost), let the personas walk through a target flow, and receive a report detailing where each persona got stuck, what they misread, and what failures the browser surfaced along the way.
The differentiating feature is the persona model itself. Rather than generic click-testing, GhostUser gives each agent a behavioral frame — Maya the Newbie misreads affordances, Dan the Buyer evaluates before committing, Alex In-A-Hurry abandons anything that adds a step. The personas generate qualitative failure narratives alongside the technical log, so you get ‘I looked for Forgot Password and closed the tab’ next to the 404 that caused it.
The tool fits product teams iterating before user research, designers who want a faster feedback loop than scheduling sessions, and developers who want automated QA layered with UX signal. It runs MIT-licensed on your infrastructure, uses your Claude API key directly, and requires no commercial account. Where it hits a ceiling: the five personas are fixed archetypes. Teams whose users skew heavily toward a specific context — accessibility needs, enterprise procurement flows, non-English interfaces — will find the built-in personas miss the most relevant failure modes and will need to extend or replace them.
The repository is structured as a monorepo with TypeScript, an examples directory, and a documented .env configuration. It integrates with Figma via a docs reference, and the vendor states it can test against localhost dev servers directly — making it usable in a pre-commit or pre-staging gate without exposing a public URL.