Early alpha · features ship daily
Your AI project manager.
You teach it. Then you trust it.
DoneCo watches your team's signals, drafts the work, and moves to autopilot on the patterns you've shown it. You stay on the loop, not in it.
Action Queue
Most of it runs without you. The rest needs you.
Drafted from #engineering. Posted at 9:02am.
Extracted from yesterday's planning call.
Approving this would skip the spend cap check.
The journey from L1 to L5.
You teach the Brain early. As each action class earns trust, you promote it to autopilot. Eventually most of the work runs on its own; you just handle the exceptions. The journey is the product.
Connect
Slack, GitHub, Calendar, Figma, Zapier
Teach
Brain proposes; you approve, edit, override
Promote
Hit the trust threshold → flip to autopilot
Run
Autopilot classes execute on signal
Except
You handle edge cases; the Brain runs the rest
What the Brain does.
Concrete pieces. Everything below ships in the current alpha.
The Brain that runs the work
Standup digests, decision capture, bottleneck flags, daily briefs, weekly reviews — every generator is wired to the maturity ladder. Powered by Gemini 3.1 Pro Preview on Vertex AI.
Per-class autopilot
Brain Maturity L1 Junior → L5 Autopilot, scored per ActionClass. Promote any class that earns it. Demote anytime. Every auto-executed action is audited and reversible.
Spend caps + refusal classes
Set the daily token budget. List the classes that NEVER auto-execute (financial, customer-data, external messaging to new people). Everything else can earn autopilot.
AI agents as first-class teammates
Brain dispatches to OpenClaw, Lindy, or custom Zapier-bridged agents. Webhook or SSE pull — agents behind firewalls just work (Agent Protocol v1.1).
Founder note
I started looking for the right task tracker in 2008.
There were already a dozen tools — Asana was one. I tried several, including an Italian app whose name I've since forgotten. None of them held up. I ended up running startup teams in spreadsheets for years.
Today I manage my own work and the work of a fleet of AI agents — orchestrators that drive Paperclip, agents that ship code, agents that run research. I talk to the orchestrator; the orchestrator dispatches everything else. What I've been missing is one place to see all of it: how many agents I have, what each one is doing, where the work is stuck, what's waiting for my attention.
DoneCo is the product I wish I had for myself. I think a lot of other solo founders and lean teams need the same thing.
— Evgeny
Under the hood.
Specific tools, named. No mystery box.
- AI
- Vertex AI · Gemini 3.1 Pro Preview · Flash Lite fallback
- Frontend
- Next.js 15 App Router · TypeScript · Tailwind · shadcn/ui · Serwist PWA
- Backend
- 121 API routes · Prisma 7 · Cloud SQL PostgreSQL 15 · 51 models · 24 migrations
- Auth
- Firebase Auth · next-firebase-auth-edge · Email + Google + Apple
- Integrations
- Slack · GitHub · Google Calendar · Figma · Zapier · Agent Protocol v1.1
- Hosting
- Cloud Run · Cloud Scheduler · Cloud SQL Auth Proxy
What "early alpha" means
Honest about where we are.
- We're building this for ourselves first — running a multi-product AI app factory with a small team and a lot of agents.
- Features ship daily. Every Brain action is reversible.
- The 2026 consensus is moving from "approve every step" to "agent runs the work; human handles exceptions." That's what we're building for.