AI coding interview practice

PraxisAI is a place to practice coding with an AI agent on real open-source issues and take-home project briefs. You work in a browser IDE, run tests and submit your changes. Graded rounds include a free score and verdict; paid plans unlock detailed reports, prompt coaching and replay of recorded events.

Pick an available GitHub issue from the SWE-bench benchmark. Your browser VS Code starts at the repository's pre-fix commit with an AI coding agent, a timer and a token budget. Inspect the code, ask the agent for changes and run local tests. On submission, separate grading tests check your patch. If the grading environment fails, the round is marked ungraded and no overall score is shown.

Signing up and browsing challenges are free. The standard allowance is 2 free full-model rounds each week. After those rounds, available credit pays for full-model usage. When the weekly allowance and credit are used, you can start up to 4 lower-quality free-model rounds a day, when available. Weekly allowances reset Monday at 00:00 UTC; the daily limit resets at 00:00 UTC. Your account shows your remaining allowance.

Free accounts see the score, verdict, test outcome and key facts from graded practice rounds. A paid plan unlocks the detailed score breakdown, prompt coach, timeline and replay where recordings are available, including detail for stored earlier rounds. Starter, Pro and Max each add the monthly agent credit shown in pricing. One-time top-ups add agent credit but do not unlock the detailed reports.

Yes. Cancel online in Billing → Manage billing (Stripe's secure portal); you never need to contact us. It takes effect at the end of the current billing period: you keep the plan until then and are not charged again. Payments are not refunded once their credit has been added, except where the law requires, but credit already on your balance stays yours to use.

Yes. Every plan can be billed annually at 20% off the monthly price, charged once a year in advance. Your plan credit is still added each month, the same amount as on the monthly plan, and unused credit stays on your balance. An annual plan renews automatically every year until you cancel, and we email you a reminder about 30 days before each renewal. You can switch between monthly and annual in Billing → Manage billing.

The standard allowance is 2 free full-model rounds each week, reset Monday at 00:00 UTC. After those, rounds are paid from your credit; a paid round needs at least $0.10 to start and is capped at what your balance covers. With no free rounds and no credit left, you can still start up to 4 free-model rounds a day on a weaker model, when available (resets at 00:00 UTC). Top up or choose a plan to keep going on the full model.

Paid rounds deduct the recorded model-provider cost multiplied by 6 from your usage credit. If the provider reports no cost, the fallback rate is $1.5 per million tokens. Round budgets and token use are shown in the workspace, and usage charges appear in your billing history. Weekly free rounds and free-model rounds do not deduct usage credit.

The report uses recorded prompts, agent responses and tool calls, edits, test runs, token use and timing. Other candidates cannot read your practice prompts or reports unless you share a report link. Public profiles are optional. Aggregate practice results can appear on the leaderboard and in race comparisons; you can opt out in settings. A company sees assessment reports for rounds it invited you to, rather than your personal practice history.

It is the same browser workspace and AI coding agent that companies use when they assess candidates on PraxisAI. Free weekly rounds and paid rounds run on the full model; free-model rounds use a weaker one. Other companies' interviews use their own tools, so we can't promise an identical setup there.

No. A company sees reports only for assessment rounds it invited you to. Your practice prompts and reports stay private unless you share a report link or publish your optional public profile. Aggregate results can appear on the leaderboard; you can opt out in settings.

The outcome checks whether the submitted changes pass the grading tests. Process dimensions use recorded evidence such as local test runs, prompt patterns and token use. They are heuristic feedback on this round, rather than proof of job performance or a guarantee that a particular prompting style will produce a better fix. Inspecting the agent's changes and running tests gives you evidence to review.

Build-from-scratch briefs with a 24-hour window that begins when you press Start inside the environment. The brief, rubric and check commands are shown up front. You work in an empty repository and submit your changes for a report, with the same free score view and paid report detail as issue practice.

The hiring side lives at its own site. Companies can send assessment links, choose tasks and review candidate submissions and process reports. Visit PraxisAI for companies, or contact us.