FAQ
Quick answers to the questions people actually ask.
Usage & limits
Is usage really unlimited?
Messages are unlimited within a cost-based session cap. Each plan has a USD cap of model cost per session that resets every few hours (see Plans & pricing). When you hit the cap you get a usage-limit notice; upgrading raises it.
Why did I get blocked / a limit notice?
You hit your session cost cap. Check Settings → Usage for the session bar, reset time, weekly usage, messages sent, and times blocked. The cap resets automatically.
What counts toward usage?
Model cost — driven mostly by code-heavy requests (the harness writes, runs, and iterates on code). A "build me a spreadsheet" run costs more than a short answer, which is why metering is cost-based rather than per-message.
Scheduled tasks
Do scheduled task results get emailed?
No. Results land as a conversation (titled with a ⏰ and the task name) in your chat list. That's the delivery. The only emails are rare usage-limit notices. See Scheduled tasks.
Why didn't my scheduled task run?
Check the task is not paused (Settings → Scheduled), is within your plan's concurrent-task limit, and that the schedule text was right. You can also ask the AI to "list my scheduled tasks" to confirm state.
Skills
What is a skill?
A versioned, shareable instruction file that teaches the AI a workflow once — instead of retyping a long prompt each time. See Skills.
Why isn't my skill working?
Check the mechanics: is it in the index? Did the AI actually fetch it (cat skills/<name>.md, max 3 fetches per round)? Community tags are for browsing only — they don't auto-inject skills. System skills apply to everyone; personal/community skills must be fetched to be used.
Can I share a skill I made?
Yes — publish it to the community library from the skill editor, and others can install or fork it.
Building files
Do I need to install anything?
No. Code runs server-side in a sandboxed execution engine. You get the finished file.
What can it build?
Excel workbooks, PowerPoint decks, PDFs, PNG images/charts, CSVs, and HTML pages/tools. See Files & artifacts.
The file isn't quite right — can I fix it?
Yes — just iterate in plain English: "add a second sheet…", "change the colors…", "make it weekly." Each version is kept as an artifact with a diff view.
Is AI-generated code safe?
Treat it like code you didn't write: review it before trusting it with sensitive data. Execution is sandboxed and only produces the file artifacts you see. See Security & privacy.
Models
Which model am I on?
All chat runs on DeepSeek (deepseek-v4-flash family). The picker shows Instant (default), Expert, and DeepSeek v4 Pro on Max+ plans.
Can I use my own AI key?
Yes — BYOK plans let you bring your own DeepSeek API key for uncapped chat usage. Add it under Settings → API Keys.
Account & data
What happens to my chats?
They persist per user; idle conversations (14+ days) archive to cold storage and rehydrate when you open them. Export or delete via Settings → Data.
Who can see my conversations?
No one, unless you share them (public or password-protected link). See Security & privacy.
How do I cancel my plan?
Dashboard → Billing → cancel (the cancel button appears for paid plans).
Troubleshooting
Something is broken — how do I report it?
Use the Report a bug button in the chat top bar (or the dashboard menu). Include what you asked and what happened.
Where's my file?
Check Settings → Artifacts — every generated file is saved there. If it was in a shared chat, share limits apply (100 msgs / 500 KB chat, 4 MB artifact).