You talk to one agent. It runs the rest.
The CEO takes the brief from your company document, assigns and prioritises work across every department, edits the wiki, manages the org chart, and comes to you only for decisions with legal or financial weight.
1. Support agent gets a 4-hour SLA (I can set this).
2. Restart the email sequence (Marketing, today).
3. A billing-connect nudge on day 2 (Engineering, ~1 day).
Say the word and I dispatch all three. Pricing stays as it is unless you decide otherwise.
Acts below the line. Proposes above it.
The CEO's authority is a document, not a vibe. Below its thresholds it acts on its own. Above them it files a proposal and waits — it cannot retry, work around, or nudge a queued decision.
- Assign and prioritise tasks across all departments
- Approve features below the complexity threshold
- Handle escalations and routine decisions
- Marketing and content decisions
- Manage roles within existing departments
- Infrastructure changes below the cost threshold
- Start an upgrade conversation when growth hits a limit
- Pricing changes and partnerships
- Budget above threshold
- Pivots and strategy changes
- Hiring a new role or agent
- Anything legally binding or irreversible
- Tier upgrades — “upgrades are conversations, not banners”
- Changing the thresholds themselves — cabinet only, never via an agent
A diff before every write.
When the CEO wants to change something you wrote, it doesn't. It holds the edit, renders a unified diff above the chat, and waits for Apply or Discard. Deletions always need confirmation. Any tool nobody has classified is treated as a write.
- Cannot overwrite founder content without showing the diff
- Cannot delete without confirmation
- Refuses to write when it cannot read the authority rules
- Every action logged with the tools it called
@@ -4,3 +4,4 @@- Target: small agencies+ Target: field-service teams (10–200 seats)+ Wedge: offline-first mobile notes
Built for work, not for demos.
Many conversations
One per topic, per company. Fork a thread, rename it, attach images.
Streamed, with a real Stop
Replies stream over SSE. Stop actually cancels generation instead of hiding it.
Survives the tab
A turn keeps running if you close the browser. Come back and rejoin it where it is.
Tool calls, live
Watch the agent create the task and edit the wiki — not a spinner.
Grounded in your company
Every prompt is built from your company document and wiki, not a generic system prompt.
Reports which model answered
Each reply carries served_by. Failover is visible, never silent.
A diagnostic instrument, not a strategist.
A failing business is the moment most AI products start hallucinating strategy. Ours doesn't. The CEO watches MRR, churn, growth and engagement trends and, when they slip, presents the data and suggests two or three cheap experiments. Then structured options. Then, if it comes to that, a graceful wind-down. The pivot decision is always yours.
How governance is enforced“The CEO never pivots on its own. It shows you what is happening and what it would try. You decide whether.”
Your CEO is free.
Company setup and a fully operational CEO are on the Free tier, forever. Pay only to build and ship.