Opportunities › Long-running AI agent management for specialist roles › Proposed app
Keepsake Desk
Keep your AI specialists' memory and voice when you move them to a newer model
- Score
- 43
- confidence 57
- Evidence
- 7
- items on the page
- Rivals
- 1
- named in sources
- First version
- 35 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 2025
- New market · AI estimate
ScreensConcept
Desktop app: Users compare agent responses, edit memory notes, and review decision history—detailed work requiring focus and keyboard interaction with dense information. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Roster
2 can upgrade5 agents
About this appAI proposal
Keepsake Desk is a workspace for people who run long-lived AI specialists, such as a Marketing Director or a Brand Director. Each agent is a saved profile: instructions, persona, memory notes, decision history, conversation log and working files. The model underneath is a setting, not the identity.
When a newer model ships, you pick an agent, choose the new model and run a trial. The agent answers a few of your past questions on the new model, and you compare them side by side with the originals. Happy? Switch. Not happy? Roll back. Nothing is lost either way.
Built for solo operators, small business owners and small teams who have put weeks into an agent and don't want to start from zero to get better reasoning.
- Tagline
- Upgrade the model. Keep the colleague.
- Voice
- CalmPlainReassuring
- Why this name
- A keepsake is something you hold on to over time. Desk is where your specialists work. Together: keep what matters.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Agents that outlive models
Persona, instructions, memory and files belong to the agent profile. Change the model without rebuilding anything.
No member yet
Trial before you switch
Replay past questions on a new model and compare answers side by side with the originals before committing.
No member yet
One-click rollback
Every switch is a versioned snapshot. Return to the earlier model and state if the new one drifts off voice.
No member yet
Decision log
Key decisions an agent helped make are kept as short entries you can read, edit and pin.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#14424ADeep Teal · primary#F6F1E7Paper · surface#1C2326Ink · text#E0902BAmber · accent#7FA38FSage · data
Typography
Aa
Keepsake Desk
ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789
Upgrade the model. Keep the colleague.
Fraunces for headlines and numbers, Figtree for everything else.
PricingAI proposal
The material gives no competitor prices. These tiers are a starting point to test, based on the opportunity's idea of pricing by agent count and history retention.
Free
Free
- 1 agent
- Basic memory view
- Export bundle
Solo
$19 a month
- Up to 5 agents
- Upgrade trials and rollback
- Decision log
- Full history retention
Team
$59 a month
- Up to 20 agents
- 5 team seats
- Shared roster
- Voice check reports
Would you pay Solo · $19 a month?
Your answer tells PWAS which prices people accept. It changes no score.
No answers yet.
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- Users invite others. Owners invite teammates to share an agent roster, review upgrade trials and edit memory together. Each invite shows a teammate the product in daily use.
- Do they become customers?
- Invited teammates are already working with the agents and could become paying seats. Test whether small teams add a second seat within a month.
- First customers
- Solo operators and small business owners on paid AI agent plans who run three or more role-based agents, such as marketing and brand directors.
- Where to reach them
- The GitHub issue thread: Reply to the commenters who asked for model switching and offer early access.
- Agent user communities: Post a short demo of an upgrade trial in forums and chat groups where users share agent setups.
- Direct outreach: Message solo founders who write publicly about their AI team and offer a free migration.
- First milestone
- Ten users migrate a real agent and five pay for Solo within two months.
Why people already pay for this
- Public claimModel improvements directly address agent needs, creating urgency to upgrade agents with new capabilities.
- VerifiedOpen GitHub issue has 43 reactions and 11 comments
- Public claimPaid users run long-running specialized AI agents as ongoing business collaborators with accumulated context and decision history.
- AI inferenceThese agents develop distinct voices, decision frameworks, and business-specific memory over weeks of investment.
- AI inference
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Model accessMulti-provider LLM API layerAgents must run on several model versions, and users bring their own keys or plans.
- This app · File storageObject storage for agent filesWorking files and exports need to be kept with each agent version.
RisksAI inference
- PlatformAgent platforms may add model switching themselves; the GitHub request could be answered natively, removing the need for a separate tool.
- TechnicalAgents on existing platforms live inside those products. Importing their history and memory may need exports or access that are not available.
- AIA new model may behave differently even with identical memory and instructions. Persona continuity is only partly preserved, so trials must be honest.
- CommercialDemand is shown by one open issue with 43 reactions. It is unproven that users will pay for a separate product.
- DataStored business decisions and files are sensitive. Users will expect strong access control and deletion.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.