Remote Workspace

Atlas's research runtime

Atlas runs safe research here. No live browser is attached — results are flagged unverified. Risky actions wait for your approval.

Mock runtime · QA only
Safety mode · always on
Atlas can browse public pages and gather research. It cannot log into financial accounts, move money, submit payments, bypass captchas, use credentials, or perform risky account actions without explicit approval. Money movement and financial account actions are always blocked.
atlas://session/idle
No active session
Send Atlas a research instruction on the right.

Recent tasks

0 total
Nothing yet. Ask Atlas to research something.

Activity log

Select a task to see its activity.

Tell Atlas what to do on the web

Browse, search, read pages, gather facts, build prospect lists. Risky actions (send, publish, buy, cancel, edit campaigns) wait for your approval.

⌘/Ctrl + Enter to send. Low-risk auto-runs; risky actions wait for approval.

Approvals

0 pending
Nothing waiting.

Prospect lists

0
Save a completed task as a list to draft outreach.

Safety

  • · Allowed automatically: web search, reading public pages, structured extraction, screenshots, CSV export, saving prospect lists.
  • · Needs approval: sending email, submitting forms, logging into accounts, changing external systems, CRM writes, payments, deletes.
  • · Mid-run risky steps pause and wait for your approval.
  • · Every record is labelled verified (live web), unverified (AI knowledge), mock (QA), or manual. These are never mixed silently.

Security & audit

Every blocked or repaired action is logged here. Use this to verify Atlas refused something rather than silently doing it.

Remote security QA: not run yet
No security events recorded yet.