FAQ
How long does a task take?
Minutes to hours, depending on the task’s scope and the thinking tier. Long runs are normal; the platform is built for them. Closing the page does not interrupt anything; come back later to check progress.
Can I add requirements mid-run?
Yes, just send a new message in the conversation. The agent folds it into the current work, but completed steps are not redone. If the new information would change earlier conclusions, it is cleaner to start a new conversation and bring the previous artifacts along from the asset library.
Why can’t I switch agents mid-conversation?
Because an agent infers “where the task stands” from the files already produced in the workspace. Swapping hands the new agent artifacts it never made, under a different skill allowlist and working method. See Core concepts.
Where do artifacts live? Do they persist after download?
Artifacts belong to the conversation that produced them and can be re-downloaded at any time. To reuse them in other conversations, click “Save to assets”. The asset library is a separate file pool under your account and does not depend on any conversation.
Where do the toolbox, agents, and skill library live?
- Toolbox (sidebar) — 26 tool cards, one workbench per tool, each doing a single, focused job: docking and scoring, formulation and phase checks, property calculators, sensory and skin feel, market dashboards, and the office file tools. Fill in parameters, run, and get a result.
- Capability Center (sidebar) — where skills, agents, and plugins are configured. Neither the skill library nor the agent gallery has a sidebar entry of its own: skills live under Capability Center → Skills, and agents under Capability Center → Agents.
How are uploaded images “understood”?
You don’t have to worry about it. The conversation model is natively multimodal: it looks at uploaded images itself, while PDF/PPT and scanned page images are extracted by the agent and understood through the image tool. The image tool is only a capability enhancement — if the external vision service goes down or times out it degrades, and the conversation is never interrupted by it.
How are credits charged? What happens when the balance runs out?
Accounts are invitation-only and come with initial credits. Conversations and toolbox tools are billed in credits, charged on success and not on failure. When the balance reaches zero, running tasks stop and new tasks are blocked (you can still sign in); everything resumes automatically once an administrator adds credits.
Will toolbox results be lost?
Yes — toolbox results are not persisted, and are lost on restart. So every workbench’s output must have a way out: tables export to CSV, images download, and structures/files download or save to the asset library. Export or save the results you want on the spot.
Why does every report have a section on limits?
Because it is required by design. The platform’s retrieval has hard limits (see Evidence sources and limits), and naming what is out of reach is far more useful than letting you trust something fabricated without knowing it.
What happens when a task fails?
The platform tries to hand you a deliverable with its gaps marked rather than nothing. Quality and integrity checks “degrade and warn” when they find problems, rather than blocking delivery.
How do I change my password or the language?
Settings → Account to change the password; Settings → Appearance to switch the language (interface and docs switch together).
How do I get another account?
There is no self-service sign-up. An administrator creates your account in the admin console with a username, email, company, and initial credits. You then receive an activation email in English (valid for 7 days); set your password on the activation page and you can sign in.
Orphic