Skip to main content

Frequently asked questions

Can I share my workspace with colleagues?

Yes. Under Settings › Users (Company Admin) you can invite team members by email and assign a role — Admin (full control), Member (standard), or Viewer (read-only). Depending on setup, a new user receives a one-time temporary password to sign in with — they should change it on first login under Settings › Account.

How often do my agents run?

That is up to you, per agent. Open the agent → Triggers tab: set a schedule (hourly, daily, weekdays, or custom cron — with a preview of the next execution times), or let it react to Telegram messages, inbound email, or webhooks. The Agents list shows each agent’s schedule; “Manual” means it only runs when someone clicks Run Now.

Who can see my data?

Your company’s data is isolated — no other company can ever access it. Inside your company, visibility follows departments, roles, and the per-record “Who sees this?” scope (see Overview & Navigation). Agents obey the same boundaries: an agent can never read data its owner cannot see.

What happens when an agent run fails?

The run is marked failed in the Runs view with the error message attached. Open the run to see the full detail — input, output, error, and a timeline of every tool call. A banner on the Agents page warns about failed runs in the last 24 hours.

Can I add or edit agents myself?

Yes — agent building is fully self-service. Use Create with AI to describe the agent in plain English, or build one from scratch. See Agents.

How do I export my data?

Every data view has an Export CSV button that exports the currently filtered rows and visible columns. For programmatic access, use the MCP API or the CLI.

What do agent runs cost?

Agent runs consume credits from your company’s balance — in-app chat is free. Your remaining balance is always visible at the bottom of the sidebar; the full picture (plans, top-ups, per-run usage log, budget caps) is under Settings › Billing (admin). See Settings.

Troubleshooting

I can’t sign in

  1. Check that you are on your company’s workspace URL and using the email address you were invited with.
  2. If you were just invited: use the one-time temporary password from your administrator, then change it after login.
  3. Forgot your password? Use the forgot password link on the login page to reset it.
  4. If none of that helps, ask your Company Admin to check your account under Settings › Users.

A page won’t load or shows an error

  1. Hard-reload: Cmd+Shift+R (Mac) or Ctrl+Shift+R (Windows).
  2. Try another browser (current Chrome or Firefox recommended).
  3. If only one page is affected, note which one and take a screenshot for your support contact.

An agent shows “failed” or has gone quiet

  1. Open Runs, click the failed run, and read the error section.
  2. If the agent simply has not run for a while: check its Triggers tab (is the schedule on? is the agent Active?) and click Run Now to test.
  3. Check the credits card in the sidebar — runs stop when the balance or the monthly budget cap is exhausted.
  4. If the error mentions a provider or API key, check Settings › API Keys.

I can’t approve something

Approvals are role-gated: mission approvals can be decided by the step’s designated approver, the mission owner, or a lead; company-level surfaces require the Company Admin role. If you believe you should have the rights, ask your admin — and try a reload first, in case it is a stale page.

Data looks outdated

  1. Check the Runs view for when the responsible agent last ran, and trigger Run Now if needed.
  2. For ingested data (e.g. shop orders), check with your admin that the external sync (e.g. Make.com) is running — rejected rows land in Failed Imports rather than disappearing silently.

Getting help

Your workspace has a support contact configured by your administrator (shown in the app’s branding). When you write in, include four things so the problem can be found fast:
  1. What happened — one or two sentences.
  2. Where — which page or which agent.
  3. A screenshot or the error message, if there is one.
  4. When — the time it happened, so the logs can be matched.