Skip to content
BotCircuits Space

Using Space Get started

Three ways to work

Quick task, team of agents, or long-running process. You just describe what you need, and Space decides how to run it.

What you ask, and what Space does

You ask for…Space runs it asExample
A quick question or lookupOne agent“Best company in Winnipeg to partner with?”
A complex task with several partsA team of agents“Simulate outreach to 100 customers.”
A recurring process that needs approvals and a recordA long-running workflowAn insurance claims check every morning

You can also take control and define a process yourself. See When you want to define it yourself.

1. A quick question

YouAgentWeb or toolsAnswer

One agent takes your request, uses whatever tools it needs, and replies. You do: type the question. Space does: everything else.

Try asking

Find the three biggest competitors in our region and list their websites.

2. A team of agents, for a complex task

YouAgentAgentAgentResult

A complex task has several parts, or needs different kinds of work, such as research, drafting and review. Space generates a team for it: a Dynamic Agent Graph. The agents pass work to each other, and the last one hands you the result. You do: describe the task. Space does: picks the agents, the order, and what each passes on.

You can also name the agents yourself if you want a say.

Try asking

Ask @ada and @bo to discuss the pricing proposal and come up with a better suggestion.

@ada prepares next week's plan, then hand over to @bo to review it.

See Agents working together and Agent orchestration.

Example: simulate customer outreach and let the agent improve

Try asking

Simulate outreach to 100 customers.

Space sees a complex task with several parts and generates a team of agents:

Outreach agent writesSimulator repliesReviewer scoresResults
  1. The outreach agent drafts personalised messages.
  2. A customer simulator replies the way different customers would.
  3. A reviewer scores what worked and what didn't.

You didn't have to ask for the improvement. It happens automatically. Agents have Self Improve switched on by default, so after the task the outreach agent reviews what happened, keeps what it learned as notes, and can refine its own instructions. The next time it runs, it starts from what the simulation taught it.

You stay in control: changes to its notes are applied, and instruction changes that need your say-so appear under Improvements, where you can accept, dismiss or revert them. See Agents that learn.

3. A process that repeats

ScheduleStepsAgentHuman gateAgentDone
  1. A schedule (or an event) starts it.
  2. Fixed steps run first: fetch data, check it, prepare it.
  3. An agent handles the part that needs judgment.
  4. The workflow pauses at the human gate until a person approves.
  5. After approval, an agent finishes the remaining work.

This is how you'd run, for instance, a daily insurance-claims review where an agent drafts a decision and a claims manager approves it. Every step is recorded.

When you want to define it yourself

Space's automatic choice is a good start. When a process matters enough to repeat exactly, define it yourself as a Managed Agent Graph, a playbook, with your own steps, agents and human gates. Or ask the copilot to turn something that worked into one.

See Playbooks, Reviews and Schedules and triggers.

Where Space fits best

  • Insurance: long-running claims and policy processes, with approvals and a full audit trail.
  • Marketing: agents draft and schedule social posts, with a human gate before anything is published.
  • Everyday starter: review your automations and list the runs that went wrong.