Skip to main content

Chat features

Quick reference for what the chat input and chat view can do.

Mentions and slash commands​

  • @ mentions a file; its content is fed to the AI as context. Type @ and start typing a filename.
  • $ mentions a skill, pulling a reusable instruction set into the turn.
  • / opens slash commands: /plan toggles plan mode, /fork branches the chat. Skills are also reachable through /.

Queue or interrupt​

Send a message while the AI is working and it queues for the next turn by default: the current task finishes, then your follow-up runs with full context. Prefer to cut in? Switch the behavior from queue to interrupt in General settings.

Fork a chat​

Branch from any response (/fork or the response menu) to explore an alternative without overwriting the original thread.

Automatic context compaction​

When a conversation exceeds roughly 80k tokens (configurable per model), Computer summarizes older messages into a rolling summary so the chat keeps fitting the model's context window. The full chat history is preserved: compaction changes what's sent to the model, not what's stored. Each turn allows up to 2048 tool iterations.

Reasoning, usage, and persistence​

  • Reasoning display: thinking models (o3, Claude, ...) show their chain of thought as an expandable section on the response.
  • Usage: every response shows token counts and timing.
  • Persistence: chats are stored per workspace in <workspace>/.cptr/chats/ as plain files you can search, move, and commit. Open the same workspace from any device and every chat is there, ready to resume.
This content is for informational purposes only and does not constitute a warranty, guarantee, or contractual commitment. Open WebUI is provided "as is." See your license for applicable terms.