opzero
Get access Sign in
Get access

opzero · company brain that spawns and coordinates.

govern agent work.

Give agents your company and product context. Run the work in a sandbox. Meet them in Slack, email, and each agent's inbox. Humans stay in the loop. Approval stays with you.

Get access

opzero product demo

Multiplayer

Agents and humans, in the same chat.

Same GC. Same thread. Agents sit next to the team in Slack, email, and each agent’s own inbox. Humans stay in the loop, company brain stays home.

Agents

  • Cursor
  • Claude
  • Codex
  • Pi
  • OpenCode
times

Humans · channels · inbox

  • Humans
  • Slack
  • Agent inbox

The brain

pre-ea

Stop re-briefing the company every morning.

The mega brain holds fat contexts so agents don’t skim. Load the map once. Wire it into the agent. Keep authority on the record when the thread moves.

every morning

Re-brief tax

You paste the same product truth into a new chat because last night’s agent forgot who you are.

session ends

Lost thread

The decision lived in a Slack scroll, a Notion page, and a half-finished plan. Nobody owns the map.

tabs everywhere

Context scatter

Sources stay siloed. Agents skim fragments. Authority never joins the receipt.

Fat contexts loaded once Never skim full map One ContextPort sources in Plug into agent same gates

Wire the brain into Cursor, Claude, Codex, OpenCode, or wherever the work already lives. Same context. Same gates.

Orchestration

Compute, spawn, and coordination.

CF Workers is our isolate adapter for sandboxed compute. Vercel and AWS are planned-compatible targets, not live-certified. Hyperagent for agent threads. Stitched to the company brain. Same OS, clear stops.

Execution worlds · isolates + planned · planned ≠ live-certified

  • CF Workers Isolate · sandbox
  • Vercel Sandbox · planned
  • AWS Sandbox · planned
  • Hyperagent Agent threads
  • Process host Local spawn
Sandbox providers
Warm isolated compute before work touches the real stack. Bound by default.
Harness requirements
Tools and checks ride with the run so the path stays inspectable.
Spawn + compute
Open a secure context, spawn agents, coordinate until the bar is met.
Cords stitch the OS
The brain plugs into orchestration through cords: plan, assign, pause, stop.
Human gate
Propose → approve → receipt. Approve in Slack or the OS. Provenance stays joinable.

Waitlist open · Get access → onboard

Get access

Sources

Connect the tools you already use.

Pull product decisions, code, docs, and design into one surface. The company brain stays home: memory and authority with op0, not scattered across tabs.

MCP · SDK · OpenAPI · GraphQL

Comms

Product

Infra

Connect

Wire into op0 so briefs, approvals, and threads land with the rest of your company context. Sign up. We’ll open the product to finish the connection.

Sign up to connect

Browse connectors

Search the connector index, then place a source where it belongs in this toolkit.

Loading connectors…

Index from the open connector registry · connect stays in op0

Company memory

The next session starts where the last one stopped.

Peek the receipt stack: this run and the ones before. Company truth, product decisions, and the stop reason stay with the work.

Illustrative workflow

run_0x7a2f ship waitlist copy
  • plan · sandbox opened · harness attached
  • authority · founder approve · target locked
  • evidence · verifier pass · receipt written
  • stop · run closed · context released
run_0x91c4 attach company context
  • plan · context scoped · sources named
  • authority · ops approve · repo locked
  • evidence · diff green · receipt written
  • stop · run closed · memory kept
run_0xb3e1 govern agent spawn
  • plan · sandbox warm · tools bound
  • authority · pending human · gate open
  • evidence · waiting on approve
  • stop · paused · awaiting approve
run_0x44d9 inspect prior receipt
  • plan · replay attached · read-only
  • authority · audit view · no mutate
  • evidence · trail complete · hashed
  • stop · inspect done · released

run_0x7a2f · ship waitlist copy

Why founders care

Stop re-briefing every session.

Spend judgment on decisions, not rediscovery. Clearer stops. Fewer restarts. More software out the door.

Get access

OpenCode · session

$ load company-brain

✓ contextport ready · 0 re-briefs

# industry (cited)

OpenAI tool schemas billed · every turn

Anthropic cache-read price · ~10% of input

# product

brain loaded once → 0 re-briefs

Slack approve → OS receipt

$ approve --in slack

✓ receipt written · run stopped

  • Re-brief the company every session Context sticks. Agents start mid-sentence.
  • Runs that never close Plan → approve → receipt → stop.
  • Approvals that vanish in threads Approve in Slack. Receipt stays in the OS.

Design systems

Design next to the product.

Tokens, typeset, and registry live beside the app. Humans and agents share the same ground.