A place where AI work actually gets finished.
Undernap is an AI team for independent professionals and teams that works on your Mac — for macOS at the relaunch, with Windows and Linux planned. It does not replace your AI accounts, it coordinates them: a goal becomes a Mission with a plan, tasks, ownership, an independent review and a result you sign off.
Seven words — and every screen reads itself
You do not have to memorise any of this. The app uses exactly these words, and they always mean the same thing.
A request with an end.
A goal in your words, plus a plan, tasks, limits and a defined close. The Mission is the bracket around everything that belongs to one request.
One step of the plan.
A bounded piece of work with a success criterion. Tasks get staffed, run in bounded sessions and leave evidence behind.
A role, not a model.
A way of working with its own quality bar — Marketing, Research, SEO & AI Visibility, Design, Engineering and Knowledge / Support. The role says how the work is done; it does not say who executes it. Content is not a role of its own; it is written within each of them.
An access you own.
A connection to an AI provider through your own account. Undernap sells no usage and re-bills nothing.
All your resources together.
The whole of your connected accounts — with their capacity, their limits and their provider family. Tasks are staffed out of the Fleet.
The independent check.
A separate instance checks the result against the goal. It is never the executor and never runs on the account that did the work.
The one outcome that counts.
The binding version with its evidence and its open points — closed by your acceptance.
What Undernap explicitly is not
-
Not a
chatbotThere is no conversation window as the product. You give goals with limits — and you see the plan before anything runs.
-
Not another
AI modelUndernap trains nothing and hosts nothing. The work happens over the providers you have connected.
-
Not a
prompt managerThis is not about managing phrasings. It is about distributing work, checking it, and carrying it to a result.
-
Not an
autopilotThe plan needs your approval, judgement calls come back to you, and the acceptance at the end is yours.
A specialist is not a provider is not a model
These three layers get blurred all the time. In Undernap they are separate — and that separation is exactly where the value of several accounts comes from.
- Specialist = the roleWhat is done, and to what standard. “Research” stays “research”, whoever executes it.
- Provider = the familyWhose infrastructure answers. Two accounts with the same provider are one family but two accounts — a review may come from the second, never from the one that did the work.
- Model = the concrete executionWhich machine takes the session. That can change without the role or the request changing meaning.
Where you spend your time in the app
Four places cover the whole flow.
- MissionsWhere the request is written — and where the plan waits for the approval you give before anything starts.
- Task WallThe state: which task is running, over which connection, what is waiting and what it is waiting on.
- ResultsThe completed Results with their evidence — and the open points that still belong to you.
- ConnectionsYour Fleet: which AI accounts are connected, what they can do and where their limits are.
What should Undernap accomplish for you?
Type or paste the outcome below. Undernap interprets it locally before anything starts.
Type or paste your Mission here
Describe the result you want—not the steps Undernap should take.
Compare our three most important competitors and show me where we stand out. Success criteria: - strengths and weaknesses per competitor - three concrete opportunities for us Boundaries: - public sources only
Local · no provider contact · Start only after approval
What should Undernap accomplish for you?
Type or paste the outcome below. Undernap interprets it locally before anything starts.
Execution brief · interpreted locally
- Desired outcome
- Compare our three most important competitors and show me where we stand out.
- Recognized success criteria
- strengths and weaknesses per competitor · three concrete opportunities for us
- Recognized boundaries
- public sources only
- Prepared structure
- 1 Tasks · 1 wave · Plan V1 after adoption
- Recommended specialist
- Research
- Open questions
- 1 · shown in the Mission plan
- Provider contacts
- 0
Local interpretation. Review the task structure and missing criteria before adoption. Model wishes do not bind execution resources.
Missions, Tasks, and Results are limited to this Project.
Understood → Plan
Review the plan
This plan was prepared locally. Adjust the Tasks, then save it as a Mission.
Desired outcome
Compare our three most important competitors and show me where we stand out.
Desired outcome
Compare our three most important competitors and show me where we stand out.
Success criteria
strengths and weaknesses per competitor · three concrete opportunities for us
Before start
You stay in control
Saving only creates a local Mission. On the next page you see the final approval and start step.
Workspace
My startup
Team
ResearchRole derived from the brief. You can change it per Task.
Usable now
Claude · Codex
Recommendation without binding; the Router decides at start. Paid execution only with your approval.
Open points
- The research asks about the outside world. Without a research connection the Specialist works from Project evidence and marks external claims UNKNOWN.
Compare our three most important competitors and show me where we stand out
One Mission, one agreed outcome.
Next step
One click approves the plan and starts exactly one bound read-only Task.
The outcome is ready
Three competitors compared: strengths, weaknesses and three concrete opportunities for us — backed by public sources.
3key findingsHow the outcome is taking shape
Trust in the outcome
Compare our three most important competitors and show me where we stand out
Three competitors compared: strengths, weaknesses and three concrete opportunities for us — backed by public sources.
Your outcome
Provider A wins on price, B on integrations, C on personal service. Open for us: an easy start for small teams, results you can trace, and support in German.
3key findingsWhat supports it
- Local verification
- VERIFIED
- Repository state
- Not checked
- Executed by
- Undernap AI team
- Provider contacts
- 1
- Limits
- public sources only
Ready is not yet acknowledged
Review the outcome, then acknowledge delivery for this exact Result.
Mission draft saved locally.
Interactive rebuild of the real interface · Founding Beta 146 in dark mode · example data · nothing runs, no provider is contacted.
Platform and operation
- macOS at the relaunch, Windows and Linux plannedAt the relaunch a native app for Macs with Apple Silicon and macOS 15 or later. Windows and Linux are planned; we will name a date once there is one. There is no version for Intel Macs.
- Local firstProject knowledge, missions and results live in your local data directory. Outbound, Undernap talks to the AI providers you connected and to web pages you ask it to analyse.
- Your credentials stay yoursConnections and keys stay on your machine. They leave it only towards the provider they belong to.
- BilingualThe app and this website exist in German and English.
See it as a flow
How it works, step by step — and a message from us at the relaunch.
Undernap for macOS (Apple Silicon, macOS 15+) is planned to relaunch on November 20, 2026. From then on it is a monthly subscription, for businesses only; your first subscription starts with a free 14-day trial. Until the relaunch we take no orders and offer no download.