AI works one person at a timeSingle-user sessions don’t scale

Get AI working with your whole team. Keep every decision yours.Multi-agent AI in one shared context. Every approval stays human.

Today each person asks their own AI, and the answers stay in their own chat. Team Cora brings your people, Cora and her AI heroes into one shared mission.Today each person runs a private assistant session, and the context dies there. Team Cora puts people, an orchestrator agent and specialist agents in one shared mission thread.

No card needed · works in your web browser · every feature on every plan
Built for teams, not solo chatsMulti-agent orchestration, one thread

One mission. Everyone on it.One context. Every agent in it.

Cora plans the work and calls in AI heroes for the specialist jobs. Your people see every step in the same chat, and they make the calls.The orchestrator plans, delegates to specialist agents and runs the tools in one visible thread. Humans stay in the loop and own the decisions.

No card needed · works in your web browser · every feature on every plan
Works while you sleep. Waits for your yes.Async runs, human approvals

The legwork gets done. The big calls stay yours.Agents run async. Approvals stay human.

Cora and the heroes work in a walled-off cloud sandbox, on a schedule if you like. The actions you choose wait on an approval card until the right person says yes.Agents execute in an isolated cloud sandbox, on a schedule if you set one. Write tools set to Ask first wait on an approval card for the human who owns the call.

No card needed · works in your web browser · every feature on every plan
Not a personal AIMulti-agent, multi-human

Team work doesn’t belong in a private chat.Team work doesn’t belong in a single-user session.

Everyone on the team has their own AI. So the work happens where nobody else can see it.Everyone runs a private assistant. So the prompts, the context and the outputs stay where nobody else can see them.

69%of respondents said most AI-assisted work happens in private sessions rather than shared spaces.Miro, The State of Intelligent Collaboration: a survey of 9,500+ leaders and employees, September to October 2026

Three things teams say
  • “I asked the AI, then pasted the answer into the team chat.”
  • “Who checked this before it went out?”
  • “We pay for AI for everyone. I can’t see what it’s done for the team.”

One chat. Your people, Cora and her heroes.One thread. Your people, the orchestrator and specialist agents.

Everyone sees the work.Shared context by default.

Questions, drafts and results are in one place, as they happen.Questions, drafts and results live in one mission thread, as they happen.

Compare with a personal AI
The old waySingle-user assistantAsk AI → Copy → Paste in email → RepeatYou become the go-between for your AI.With Team CoraShared multi-agent workspaceMeganDanielLeahCoraKitOne mission. Everyone who owns the work is in it.

The right person makes each call.Human-in-the-loop, routed by role.

Cora sends every decision to the person who owns it.Cora routes each decision to the person who owns it.

See how Cora works

It carries on while you’re off.Async by design.

Scheduled work runs in the cloud overnight. The results are waiting in the morning.Scheduled runs execute in the cloud sandbox overnight. The artifacts are waiting in the morning.

How the sandbox works
AI with manners. It asks first.Human-in-the-loop by default

Meet Cora. She coordinates. You decide.Meet Cora. She orchestrates. Humans approve.

We know what it’s like when everyone has their own AI and the team is no better off. So we built an AI that works in the open, with the whole team, and asks first wherever you choose.We know what it’s like when everyone has a private assistant and the team is no better off. So we built an orchestrator that works in the open, in shared context, and waits for approval wherever you set Ask first.

“Tell me what you want done. I’ll coordinate the rest. You make the calls.”“Give me the objective. I’ll plan, delegate and report back.”

Has never said “oops, already sent”.

  • She asks first.Write tools ask first.A tool set to Ask first waits for an approval card before it changes anything. Your admins, or the person who added the connection, choose which ones.A write tool set to Ask first waits on an approval card. Admins set this per tool.
  • She never sees your passwords.Credentials never enter the model’s context.Logins stay in a vault. The AI uses them without reading them.Secrets stay in a vault. Agents refer to a key by its name; its value never reaches a prompt.
  • She works in her own walled-off space.Isolated execution.A cloud sandbox, not anyone’s laptop.An isolated cloud sandbox, never anyone’s device.
How Cora keeps things safe
Sydney team

Invested in AI, but it never stuck?AI pilots stalled? We get agents into production.

Built and supported in Sydney. We set it up with you, write skills for your work and build custom connectors for bespoke and legacy systems.Built and supported in Sydney. Our team writes domain skills, builds custom connectors and MCP servers for bespoke and legacy systems, runs evals and tunes token use.

How a mission runsThe orchestration loop

Three steps to your first missionThree stages to your first mission

  1. Tell Cora what you want doneObjective in
  2. Approve the team she suggestsApprove the proposed team
  3. Make the callsHumans approve

Tell Cora what you want done.Give the orchestrator a goal.

Write the goal in your own words.A plain-language objective becomes the orchestrator’s context.

Approve the team she suggests.Approve the team the orchestrator proposes.

Your people and the AI heroes for the job. Untick anyone.People by role and specialist agents by capability. Untick any line before you approve.

Make the calls.Stay in the loop.

The heroes do the legwork. Anything set to Ask first waits for your yes.Specialist agents run the tasks. Write tools on Ask first wait on your approval card.

Meet the heroes
Inside a heroAgent anatomy

Build a hero once. Your team calls it in.Persistent specialist agents, shaped by your team

Your team decides what each hero knows, what it may use and who signs off its work. It stays on your team, ready for the missions you open it to.People stay in the loop: they choose each agent’s skills and connections, and the people who approve its work. Agents persist across missions and are called by handle.

@kitChecks a page before launchSpecialist agent: page QA
  • SkillsWhat it knows how to do, chosen and approved by your team.Domain skills: instructions and scripts for repeatable, predictable results. An admin approves each one.
  • ConnectionsConnectors and MCPThe tools and accounts switched on for its mission, and nothing else.Connectors and MCP servers switched on for its mission. Write tools on Ask first wait for approval.
  • MemoryWhat it has learned, kept as plain Markdown your team can read.Persistent mission memory, kept as Markdown your team can read.
  • A walled-off workspaceSandboxYour organisation’s sandbox in the cloud, with a private folder for each call.Runs in your organisation’s isolated sandbox, with a private scratch folder per call.
  • AI modelModelRuns on Claude, from Anthropic, matched to each job.Tasks are routed across Anthropic’s Claude Haiku, Sonnet and Opus.Choosing the model per hero: coming soonPer-agent model choice: coming soon
  • Who decidesHuman approversThe people who must say yes, by role.Role-based approvers: the humans in the loop.

How Team Cora works

See how Cora works

Team Cora is a shared workspace in your web browser. You open a project, which we call a mission, and say what you want done. Cora, the mission coordinator, plans the work and suggests who should be on it: the people on your team, and the AI specialists we call heroes. Each hero has its own job, skills and tools, set up once and shared with everyone. The work happens in one chat that the whole team can see. When something needs a decision, or a tool set to Ask first is about to change something, Cora asks the person who owns that call and waits. What the team learns is kept as plain documents you can read and download.Team Cora is a shared workspace that runs in the browser. You open a mission and state the objective. Cora, the orchestrator agent, plans the work and proposes the team: people by role, and the specialist agents we call heroes. Each hero has its own role, skills and connectors, configured once and shared across the organisation. The work happens in one mission thread that every member can read. When a decision is needed, or a write tool on Ask first is about to change something in a connected system, Cora routes an approval card to the person who owns that call and waits. What the team learns persists as Markdown artifacts you can read and download.

Questions teams ask

Will it send something without asking?Can an agent act without approval?
Not when it’s set to Ask first. A tool on Ask first waits for a person’s approval before it changes anything, and your admins, or the person who added the connection, choose which tools are on Ask first.Not on Ask first. A write tool on Ask first waits on a person’s approval card; admins choose which tools are on Ask first.
Can it see our passwords?Do credentials reach the model?
No. Logins stay in a vault, and the AI never reads them.Secrets stay in a vault. Agents refer to a key by its name and never read its value.
Does it run on my laptop?Is there any local execution?
No. It works in a walled-off sandbox in the cloud, with no way into your computer.Agents run in an isolated cloud sandbox, with no path to your devices.
Is it only for marketing teams?Is it tied to one domain?
No. Any team, any mission.Any domain, any mission. See seventeen examples.
Do we need to be technical?Do we need engineers to run it?
No. You write in plain English, and our Sydney team sets it up with you.Objectives are plain language, and our Sydney team handles set-up, skills and connectors with you.
We already have an AI chat for everyone. Why this?We already have a single-user assistant for everyone. Why this?
Keep it for your own work. Team Cora is for the work you do together.Keep it for single-user work. Team Cora is the shared, multi-agent workspace for the work you do together.
Try it now

Pick a first mission

Any team, any mission. Choose one and see how Cora would run it with your people.Any domain. Pick one to see the plan, the human approvers, the agents involved and the artifacts it leaves behind.

Growth and launches

Sales and customers

Operations

Finance and legal

People and leadership

Cora · mission coordinatororchestrator agent
Here’s how we’d run launch a product:
  1. You set the launch date and who signs off what
  2. Cora writes the plan and drafts the announcement
  3. Kit checks the landing page before it goes live
  4. Cora asks the budget owner before any ad spend starts
Who decidesHuman approvers
Team leadBudget owner
HeroesAgents
Kit
SavedArtifacts
Launch plan.mdAnnouncement.md
Start this mission free

The heroes named here are examples. Your team builds its own, or our Sydney team builds them with you. Cora prepares and checks; she does not give legal, tax or financial advice, and your people and your advisers make those calls.The agents named here are examples. Your team builds its own, or our Sydney team builds them with you. Agents prepare and check; they do not give legal, tax or financial advice, and your people and your advisers make those calls.

Every plan has everything

Every connector, any MCP server, the cloud sandbox, vaults, schedules and heroes. Plans differ only in AI usage, people and support.Every connector, any MCP server, the sandbox, the vault, schedules and specialist agents on every plan. Plans differ only in AI usage, people and support.

Free for 14 days, then fromUS$49a month for 5 people

See pricing
Cora asks

Ready to start your first mission?Ready to run your first mission?

It’s your call. No card needed.Your approval. No card needed.