The code workspace in guest mode
How the code workspace behaves in signed-out chats, and the rate limits that apply.
You can use the platform without an account, and anonymous ("guest") chats can still run code in the code workspace. Guest sessions run through a separate, isolated proxy path that keeps them rate-limited and credential-scoped, so they are a lower-privilege slice of the full experience rather than a full account. This page explains what changes when you are not signed in and why some things are throttled. To lift the limits, sign in.
How guest workspaces are isolated
Guest workspace requests route through a dedicated guest proxy instead of the direct tunnel that authenticated sessions use. That proxy:
- Runs the workspace against a guest-scoped session keyed to an anonymous guest ID and the chat, so guest work is isolated from other guests and from signed-in users.
- Requires a valid guest ID on every call.
- Applies per-guest burst limits to workspace runs, on top of the create limits below.
- Allows a smaller set of operations: shell commands, Python, R and notebooks; reading, editing, listing and saving files; installing extra packages; and fetching public data or repositories. Octave, starting compute jobs from workspace code, multi-step hand-offs, lab-automation protocols and native figure rendering (gnuplot, Circos) need a signed-in account.
The create rate limit
The most visible guest limit is on starting a workspace. Each guest chat has a cap on workspace starts within a 10-minute window, and each guest client has a daily cap across chats. When a start is refused, the server returns a retry-after time, and the chat cannot run code for 10 minutes. Ivy tells you the workspace is temporarily unavailable instead of failing silently.
This is why a guest chat whose workspace keeps stopping and restarting can stall: each cold start counts against the cap. Signed-in sessions do not have this cooldown.
What still works
Guest workspaces support the core interactive workflow — running shell commands and Python, reading and listing files, and saving results — which covers most analysis a guest chat will ask for. The limits you will actually feel are the create-rate cooldown above and the ephemeral nature of the session: guest work is not tied to a persistent account, so there is no long-term job history or durable project storage the way a signed-in account has.
Signing in removes the friction
An authenticated session drops the guest cooldown, connects the workspace to your persistent job history and projects, and gives you the full workspace lifecycle described in the overview.
Read next
Was this page helpful?
The code workspace
Every chat has a private Linux workspace where Ivy writes and runs Python, R, Julia, Octave, notebooks and bioinformatics command-line tools against your files and results.
Code workspace troubleshooting
Fixes for timeouts, lost files, missing packages, and stalled downloads in the chat's code workspace.