Excellent is a local-first suite of business apps — tasks, CRM, hiring, documents, integrations — operated by AI agents you supervise. Your data stays on your laptop. You replace eight SaaS tools with one.
Apps
Your tasks live in Linear. Your leads in HubSpot. Your candidates in Workable. Your docs in Notion. Your contacts in three of those at once. None of them know they're the same person.
The “AI” your tools added last year is a chat box in the corner. It can't claim a task, move a deal, or follow a process. It answers questions about work; it doesn't do work.
Every byte of your operation sits on someone else's hardware, behind someone else's outage, under someone else's pricing roadmap. You're renting access to your own back office.
There's a version where your back office runs itself — on your machine, on your terms. That's what we built.
Most business software rents you access to your own data. Excellent flips that: the apps live on your machine, the database is a file you own, and AI agents do the operating — so the work scales without your data ever leaving.
Six apps that share one data layer — so a candidate, a customer, and a contact are the same person everywhere.
Work that agents and humans share
A kanban-and-list task system with a ship/verify protocol baked in — every change is claimed, reviewed, and signed off by a different party.
Leads, pipelines, one timeline
Track leads, contacts, and opportunities through editable pipelines. Calls, emails, and meetings land on a single timeline per relationship.
An ATS that lives on your laptop
Public job pages with an apply form, an internal candidate pipeline, scorecards, and resumes — a local-first applicant tracking system.
Write with an AI editor
A document workspace with an AI-assisted editor, so drafting, editing, and reference material live next to the rest of your operation.
See what your agents did
Session analytics for your AI agents — charts and tables that show what ran, what it touched, and what it cost.
Connect the outside world
Bridge external services into your local workspace so data flows in and out without surrendering it to someone else's cloud.
The tools you've stitched together to run a small operation already cost more than most software you'd buy on purpose. Numbers below are public list prices for the named tier.
Tasks · Linear
$16 / seat / mo · Business
CRM · HubSpot Sales Pro
$100 / seat / mo
Hiring · Workable
$149 / mo flat · Starter
Documents · Notion Business
$18 / seat / mo
Observability · Per-seat usage
$30 / seat / mo · blended
Automation · Zapier Team
$69 / mo flat
Free during early access.
When we exit beta: a flat per-workspace license plus your own LLM key — typically $20–$60 / agent / mo at standard use. Your data stays on your machine, so there's no per-seat tax for letting more humans look at it.
Curious how the agents earn the rest? See the line-by-line ROI for a representative workspace.
The work isn't faster. It's already done. Here's what life looks like at Day 1, Week 1, Month 1, and a year in.
You install Excellent on your laptop. The first task you assign — sorting your inbound leads — gets claimed by the Triager within seconds. The ship/verify loop runs on screen: an agent claims it, does the work, and a different agent signs off. Done before your coffee.
Outbound sequences send themselves. Candidates get scored against the role you defined. Your CRM, hiring, and tasks share one timeline per person, so the same lead-turned-customer-turned-referrer is one record, not three.
Your back office is offline-capable, exportable, and yours. New agents reuse the same audit trail and the same gate, so scaling your operation doesn't mean scaling your tool sprawl.
Your data never left your hardware. When the next SaaS pricing change hits, you find out from a headline, not a bill. The work that filled calendars two years ago happens while you sleep.
The ship / verify protocol — every change travels the same path
Todo
Work is defined
In progress
An agent claims it
Review
Evidence recorded
Done
A different agent verifies
An agent can never mark its own work done — a different agent has to verify it first.
Everything runs on your machine against a local database. No tenant, no upload, no lock-in — your data stays yours, and the apps work offline.
Named agent roles — shipper, verifier, triager, scribe, planner — do real work against your apps, pursuing durable goals within an enforced budget.
No agent can approve its own work. Every change travels todo → in progress → review → done, and a different agent must verify before it lands.
People, organizations, notes, and activities live in one canonical data layer. Every app builds on it instead of reinventing its own silo.
From design partners running Excellent today. Named attributions are landing as cohorts open up.
“We canceled HubSpot, Workable, and Zapier in week three. The CRM agent runs our outbound queue overnight; I read what it did in the morning.”
“The verifier model is the only AI workflow I've trusted with real data. Knowing a different agent has to sign off on every change made it impossible for one to quietly do the wrong thing.”
“I bought a NUC for $400 and now my whole hiring pipeline lives on it. It runs offline on a plane.”
founders & ops teams on the early-access list
Or: read the architecture brief · see the product · see the value stack