Task and objectives
The problem, objectives, rules and submission format, written around your workflow.
Test AI approaches on your team’s work.
Turn your day-to-day work into scored challenges, compare approaches side by side on shared test cases, and keep the best solutions as skills in your company registry.
Built for
Teams applying AI to their own work
What you get
Scored challenges, leaderboards and reusable skills
Get started
Guided program
We scope the program with you, one workflow at a time.
Why it matters
Compare approaches on shared test cases.
Shared test cases give your team a way to compare approaches to the same task.
Ad-hoc experiments
- Separate experiments without shared objectives
- No common test cases to compare one approach with another
- Prompts and workflows difficult to share
With a challenge program
- Each workflow defined as a challenge with test cases and a rubric
- Attempts scored on shared test cases, on one leaderboard
- Selected solutions shared to your company skill registry
How it works
How a program runs.
Scope
We work with you to pick the workflows, their owners and their data limits, and agree deliverables, timing and fee.
Draft
Each workflow becomes a private company challenge, drafted in Versalist and published by your admin when it’s ready.
Attempt
Run attempts in the web app and review scores against the test cases and rubric. For private challenges, project submissions are limited to the challenge creator.
Keep
Share the solutions you want to keep to your company skill registry, where your review policy applies.
Features
Define and score a challenge.
Private test cases
Inputs with expected outputs, kept private by default so nobody can tune to the answers.
Weighted rubric
Graded or pass/fail dimensions with weights, locked once others have scored runs so results stay fair.
Consistent judging
The same Versalist judge scores every attempt, and each challenge sets which models and API keys a run may use.
Benefits
Compare and reuse approaches.
Decisions backed by scores
Choose which AI approach to roll out based on how it scored on your own work, not on a demo.
Share useful approaches
Share selected solutions as skills for your team to review and use.
Reuse test cases
Run later approaches against the same challenge and rubric.
Integrations
Works with the models and tools your team already uses.
- Model providers
- OpenAI
- Anthropic
- Your keys
- Bring your own provider API keys
- Public challenges also run from
- Versalist CLI
- MCP server
- Claude Code
- GitHub Copilot
- Codex
- Cursor
- Antigravity
- OpenCode
- JetBrains Junie
Resources
Explore public challenges and the docs.
Public challenges
Browse published challenges with their brief, rubric and public test cases.
Browse challengesChallenge format
What a challenge contains and how attempts are scored.
Read the challenge docsVersalist CLI
Start and submit public challenges from a terminal, or connect an agent over MCP.
Read the CLI reference
Security
Company challenges are private by default.
Access follows your company roles, from the first draft to the final leaderboard.
Company-only access
Company challenges are private to your members by default, and only an admin can open one up.
Admin-controlled publishing
Your company admins create and publish company challenges.
Private test cases
Test cases stay private unless you mark them public, and private cases are never sent to participants.
Secure credential storage
Model-provider keys are stored per user in Google Secret Manager.
Pricing
Know the cost before you start.
Fixed fee, agreed before the program starts.
Deliverables and timing are agreed with you during scoping.
FAQ
Frequently asked questions
How long is a program?
We agree the length with you during scoping, based on how many workflows you want to cover.
What do we need to bring?
An owner for each workflow, source material you can use, and time to review the challenge drafts.
Who works on the challenges?
Your own team, in the Versalist web app. Company challenges are private to your members by default.
Does every solution become a skill?
Only the ones you choose. You decide which solutions to keep, and your review policy applies.
Can we try a public challenge first?
Yes. Anyone can browse public challenges. Running one needs a free account, which our team activates after you sign up.