Prompts, agents, workflows, MCP tools and skills
Your prompts are already software.
They live in docs, chat threads and code comments, untested and unversioned, quietly breaking when a model changes. Prompt Verse gives them a real home: compose them from typed pieces, version them with branches and merges, prove changes with evals. Your own AI runs them.
Request access See how it works
Pre-beta, invite-only. No built-in AI and no credits to buy: Claude, ChatGPT, Grok or any MCP assistant does the running. The windows on this site show sample prompts and sample numbers.
Works with the AI you already useClaudeChatGPTGrokany MCP assistant
What it does
Compose. Version. Prove. Publish.
Five kinds of AI asset share one lifecycle, so a prompt, an agent and a workflow are written, reviewed and shipped the same way.
Compose
Build prompts from typed pieces: sections, modules, blocks and wrappers such as role, constraint and format. Mark the {{variables}}.
The libraryVersion
Commit, branch, tag and merge, for prompts, agents and workflows. Go back to any version and see what changed.
VersionsProve
An eval is a set of cases and a way to score them. Compare two versions and read the difference as a number.
EvalsConnect
Your own AI finds, fetches and improves your prompts over MCP, and runs them on its own subscription.
Your AI
Versions
Branches, tags and a real merge.
Two people take a prompt in different directions and bring it back together. Commit the working copy, branch to try an idea, tag the version that shipped, merge when it works. A conflict comes back to you with both sides, not as a silent overwrite.
- Commit, branch, checkout, merge, tag
- Prompts, agents and workflows all use the same history
- Restore an older version by branching from it
Evals
Make “better” a number.
An eval is a list of cases (an input and what you expect) plus a method to score them: exact match, contains, structure, a judge, or a mix. Run two versions on the same cases and the difference is no longer an opinion.
- Five scoring methods, from exact to judge
- A threshold you choose, 0.8 by default
- Your own AI can read the cases and score the answers
A door for your AI
No copy and paste. Your AI asks for it.
Prompt Verse runs no model. Your AI connects once over MCP, finds a prompt in your library, asks for it rendered with the variables filled, and runs it itself. You can still copy any prompt from the app when you prefer to.
- One connection for Claude, ChatGPT, Grok and other MCP assistants
- Search, read, render, update and commit from your AI
- Through Geenius OS or with the app’s own MCP address
In numbers
Facts you can check, not testimonials.
Prompt Verse is in pre-beta and has no public customers yet. These are things you can look up in the app and the docs today.
- 5kinds of AI asset on one lifecycle
- 11render formats, from Markdown to Prompty
- 5ways to score an eval
- 0credits or tokens you buy from us to run a prompt
Honest labels
What works today.
We say what is live, what is in pre-beta and what is only planned.
- Pre-betaThe library, versions, evals and sharing
Built and in the app now. The pre-beta is invite-only.
- Pre-betaConnecting your AI over MCP
Works through Geenius OS and through the app’s own MCP address.
- Pre-betaThe marketplace
Listing, installing and withdrawing work. Paid checkout is mocked until payments open, so nothing is charged.
- PlannedSelling prompt runs without showing the prompt
Planned: creators sell runs and the buyer never sees the prompt. Not live, no date yet.
- Not offeredA built-in AI
There is none during the pre-beta. Your own AI runs your prompts.
Who uses it
For people who ship with AI every day.
Developers
Branch a prompt, prove it with an eval, merge it, and let your services fetch the tagged version.
See the setupCreators
Keep your best prompts, agents and skills in one place, in every format your tools need.
See the setupTeams
Assets belong to a workspace with members and roles. Private by default; public only when you publish.
See the setupSupport and operations
A reply prompt with a role, a constraint and variables for the order. Fix the wording once and everyone gets it.
See the setup
Pricing
Three plans. No free tier.
Per month, in euros. Included in the matching Geenius OS plan. Payments are in test mode during the pre-beta.
- €9Pro: the workspace, 50 prompts, 25 workflows, 25 agents
- €19Premium: double the room, carries the AI features
- €49Max: no counts, carries the AI features
From the blog
Notes on prompts as software.
Your prompts are already software, so give them version control
A prompt that decides what your product says is code in everything but name. Four things software teams do that prompt owners usually skip.
How to write a prompt with variables that does not break
Turn a prompt you retype into a template. How to choose variable names, what to keep out of them, and how to find the missing ones before your AI does.
Test a prompt with five cases and no tooling
A practical way to check whether a changed prompt is better: pick five inputs, write what a good answer contains, score both versions and keep the winner.
Questions
Good questions.
What is Prompt Verse?
A workspace for AI assets: prompts, agents, workflows, MCP tools and skills. You compose them from typed pieces, version them, test them with evals and share them. Your own AI runs them. It is an app of Geenius OS and also works on its own.
Which AI runs my prompts?
Yours. Prompt Verse has no built-in AI in the pre-beta. Connect Claude, ChatGPT, Grok or any MCP assistant and it reads, runs and improves your prompts on its own subscription. Composing, versioning, sharing and publishing need no AI at all.
Do I need Geenius OS?
No. Prompt Verse is a standalone app with its own account and plans. A paid Geenius OS plan includes the matching Prompt Verse plan, and you sign in with the same account.
Is Prompt Verse ready to use?
It is in pre-beta: invite-only, with payments in test mode. The parts described on this site are built and in the app. We let people in a few at a time, so request access and we write when your place is ready.
Which formats can I render a prompt in?
Eleven: Markdown, text, XML, JSON, YAML, HTML, CSV, TOML, JSONL, chat messages and Prompty.
Is “better” an opinion here?
No. An eval scores every case from 0 to 1 and compares the result against a threshold you set, 0.8 by default. Run two versions on the same cases and the difference is a number.
Treat your prompts like software.
Prompt Verse lets people in a few at a time. Ask for a place and we write when yours is ready.