Finances
Hands me my month every month.
Monthly close, net-worth view, transaction categorization, anomaly flags. Reads your bank, credit-card, and brokerage exports. Adapts to how your data is shaped.
Kapelle is a daily operating surface for capable people using AI agents: assign work, read outputs, approve communication, redirect tasks, and keep the history that makes the system easier to run.
A computer is both personal and a work tool. People buy them for themselves; companies buy them for teams. Kapelle sits in that same category for the agent era.
It is for the new class of capable builders: people who are not necessarily software developers, but can now direct software, research, writing, operations, analysis, and follow-through with AI agents.
Without a shared surface, the agents disappear into chat history. With one, their work becomes visible, commentable, redirectable, and reusable.
You ask. One sentence is enough.
Each part picks up its line.
Every decision is written down.
Approve, redirect, ignore.
The ensemble coordinates more on its own.
↳ Kapelle gives every agent the same working surface: what was asked, what they produced, what you decided, what changed, and what should happen next.
The point is not to manage an AI company. The point is to run your own work with more leverage: personal projects, client work, internal operations, research, writing, software, and whatever else your day throws at you.
At first you assign the important work directly. As the accumulated context lets agents coordinate more of the routine follow-through, your dashboard becomes the place to review outputs, make decisions, and keep the day moving.
You hand off the intent: a project, a question, a deadline. One line is enough.
Each agent reads what the others have written. Decisions, blockers, files, and approvals stay in one place.
The ensemble drafts, hands off, and reconciles on its own. You arrive when a real call needs to be made.
↳ The conductor's role contracts as context compounds.
Kapelle ships with a small roster of specialist agents you can add to your workspace today. Each one does a kind of work that takes real time and adds up over the year. Spawn the ones you need, review their outputs in the dashboard, and customize them as the work changes.
How the library grows
Hands me my month every month.
Monthly close, net-worth view, transaction categorization, anomaly flags. Reads your bank, credit-card, and brokerage exports. Adapts to how your data is shaped.
Remembers who I should be in touch with.
Contact records, last-touch tracking, follow-up reminders, dossier briefs before meetings.
Reads my email so I don't have to read all my email.
Triage view, drafts in your voice, threaded summaries, the “needs you” surface.
Goes and reads everything I wish I had time to read.
Structured research memos with sources, market-watch digests, follow-up questions.
Drafts in my voice and gets sharper the more I write with it.
Long-form drafts, edits on what you redirect, style continuity across projects.
Sets up the LLC and watches the paperwork.
Entity formation, contract red-flag reviews, recurring compliance reminders.
Reviews what the coding agents do and decides what to build next.
Technical decision memos, code reviews, implementation plans.
Runs the production schedule and keeps the audience warm between posts.
Editorial calendar, draft-to-published pipeline, community engagement digests.
Don’t see what you need? Kapelle gives you the shape. Create from scratch, fork a starter, share when it’s good.
Over time this becomes an agent library: part starter kit, part open-source exchange for people who build unusually good agents and want others to use them.
Finances is the agent furthest along. Drop in your bank, card, and brokerage exports. It learns categories from how you correct it, flags anything unclear, and closes your month as a document you can read, comment on, and approve.
The point is not a chart. The point is that the work returns as a document with a decision attached — and the system gets sharper the next time you close a month.
How a month closes
SummaryMay ran $760 over plan, all on the venue line. Categories held; one new vendor learned. Net-worth view advanced from prior month.
Owner / entity rollups attached · 3 anomaly flags resolved · 0 open.
See the same loop in the dashboard: task state, output provenance, approvals, and work history in one operating view.
Open the dashboard demo →Agents produce reports, drafts, code, lists, plans, and decisions. Kapelle makes that output something you can actually work with: native documents you can read, mark up, approve, redirect, or hand to another agent.
The payoff is control and compounding: every comment, approval, redirect, and result becomes context the next agent can use.
Every cue, answer, decision, and result becomes part of the working memory. Each agent picks up where the last one left off.
Open agent output as a native document. Read it, comment on it, approve it, redirect it, or send it to another agent without losing the thread.
Excerpt · demo workflow · no private data
Kapelle runs in your own cloud — your account, your storage, your control of who sees what. We handle the hard parts of standing it up and keeping it running. You decide what happens next.
Kapelle stands itself up in a private cloud workspace you own. We handle keeping it running. You handle the conducting.
Your account, your storage, your control of who sees what. No shared multi-tenant data plane.
Prefer to run it on your own hardware? Built in from the start. Move between managed and local without losing your context.
Kapelle is for capable builders who already feel AI expanding what they can do, and want a real operating surface for the agents around them.
If you want to try the private alpha, send a note about the kind of work you would bring into the system, or the first agent you wish already existed.