Installing a team
How to install pre-built AI agents and run it from your workspace — no API calls or terminal required.
What this guide covers
You install a team from the workspace Library and run it from your workspace. You supply the credentials the agents need (API keys, config values), approve any records that need human review, and monitor how runs are going.
1. Find a team
Open the Library panel from the sidebar in your workspace and search for a team. You can filter by category (customer support, sales, research, etc.) or search by name or description.
The Library only shows published teams — drafts are never visible to installers. What you can see depends on the team's visibility:
- Public — visible to everyone in the Library
- Team — visible only to members of the author's organisation. If a colleague shares a link to a team and you can't find it in search, it may be set to Team visibility.
- Private — not listed anywhere; only the author can see it
Before installing, also check whether the team is forkable. Some teams allow you to copy them into your own workspace and customise them freely — others are locked and run as-is. You'll see this on the team page before you commit to installing.
2. Review the team page
Before installing, take a few minutes to read the team page:
- Agents — who the agents are and what each one does
- Required secrets and variables — the API keys and config values you'll need to provide. Collect these before starting — you won't be able to skip them during install.
- Access policy — what external services each agent is allowed to contact. A team with a declared list of allowed hosts is more predictable than one with no policy.
- Human gates — whether any steps pause for your approval before continuing
- Pricing — per-run, per-1K tokens, or subscription
- Forkable — whether the team author allows forking. If it is, you can copy the team into your own workspace and customise it. If forking is off, you run the team as-is.
- Version history — how actively the team is maintained and whether breaking changes have been communicated clearly
3. Check governance requirements
If your organisation has AI governance enabled, installing a new team requires approval from an org admin. You'll see a notice on the team page if this applies — submit the request and your org admin will approve or decline. You'll be notified of their decision.
Governance approval is a one-time step per team, not per run.
4. Install the team
Click the install button on the team page. You'll be taken through a short setup form — paste in your API keys and config values. Secrets are encrypted immediately and never shown again. Once setup is complete you'll land in your workspace with the team's room thread open.
5. Run the team
In the room thread, press Shift+R to trigger a run. You can optionally type a short description or task first — this seeds the first agent's prompt with extra context.
A run card appears inline in the thread and shows the status as it progresses — "Running…" with a live token counter, then once complete a per-agent breakdown and output preview. You can leave the page and come back — the run continues in the background.
6. Review gate approvals
If the team has human review gates, the run pauses and an amber card appears inline in the room thread showing the content waiting for review. Click Approve & continue to release the gate and let the run proceed. If you need to read the full output before deciding, click the maximize icon in the gate card header to open it in a fullscreen reader.
If you want to decline or flag an issue, describe it in the thread — the assistant can help re-run the step with corrected context or halt the run.
7. Set a schedule (optional)
After your first run, a schedule card appears directly in the room thread. Pick one of the presets:
- Every weekday 9am — runs Monday to Friday each morning
- Every Monday 9am — runs once a week to start the week
- Daily at noon — runs every day at midday
- Every hour — runs continuously on the hour
- Custom — set your own frequency, time, and timezone
To change the schedule later, describe it in the thread — "run this every Monday morning" — and the schedule card reappears.
8. Monitor runs
Your agents appear in the sidebar of your workspace with a status dot — green when live and running successfully, red if the last run failed. The age of the last run ("Run 2 hours ago") is shown beneath the team name.
Open the room thread to see recent run cards — each shows the overall result, per-agent breakdown, and any error messages. If a run fails, the run card tells you where it failed. Fix the issue and press Shift+R to retry.
For deeper history — token counts, per-run costs, filtering across all teams — open the Observability tab in your workspace.
Managing your install
Rotating secrets — If an API key expires or is compromised, an Update credentials button appears on the failed run card in the thread. Click it to re-enter the affected values — changes take effect on the next run.
Multiple installs — You can install the same team more than once — for example, a staging install and a production install with different secrets and schedules. Each install is independent.
Upgrades — When the team author publishes a new version, you may be notified in your workspace. Your existing install keeps running on the current version — you choose when to upgrade.
Customising a team — If the team allows forking, open it in the Library and click Fork & edit. Envelope copies the team into your workspace as a private draft — all agents, prompts, and settings intact — and you can edit it freely from there. If forking is disabled by the author, you can run the team as-is but cannot modify it.
Removing a team — Right-click the team name in the sidebar and select Delete from the menu. This removes the team from your workspace and stops all future scheduled runs.