TUI Guide
The TUI is the most complete way to use Grok Code. It is where setup, guided workflows, staged review, and session continuity make the most sense.
Launching The TUI
grokIf xAI is not configured yet, start with:
/connectHow The TUI Is Organized
The TUI is built around lanes and workflows, not around a long list of buttons.
The Main Lanes
Work: the primary chat and coding laneGate: review staged change artifacts before landing themResearch: search, inspect, pin, and package evidenceRecall: reopen saved sessions and branchesControl: setup, readiness, permissions, MCP, browser status, and verification status
Use /deck when you are unsure where to go.
The Fastest Navigation Shortcuts
ctrl+p: command palettectrl+g: permission modeshift+tab: plan mode togglef6: mouse scroll or selection mode/deck: lane chooser/preferences: settings modal
Work Lane
Work is where you ask Grok Code to do things.
Use it for:
- coding tasks
- reviews
- refactors
- interactive follow-up
- task modes such as
/quick,/deep,/persist, and/qa
Important behavior:
- TUI edits usually do not land immediately
- Grok Code stages reviewable artifacts and sends them to
Gate - this is different from CLI
runandexec, which apply successful edits live
Useful commands here:
/model/profile/steer/followup/search/mission
Gate Lane
Gate is the governed review surface.
Use it when Grok Code staged changes for review.
Core actions:
/gate: open the full lane/gate drawer: show the compact drawer insideWork/gate apply: land the latest or named staged artifact/gate reject: discard the latest or named staged artifact/gate rollback: restore the latest or named rollback checkpoint
What to expect:
Gatestays empty until a run produces a staged artifact- applying a change can create a rollback checkpoint
- rollback only helps if a checkpoint exists
See Governance And Verification for the details.
Research Lane
Research is where you collect evidence and feed it back into active work.
Typical flow:
- run
/research - inspect the saved result set
- use
/research pinfor anything you want kept in context - use
/context buildwhen you want a bounded package of session context plus pinned evidence
What to expect:
- research can still work even if browser-backed capture is unavailable
- MCP and browser capture improve the result when configured
Recall Lane
Recall is for saved work, not for new work.
Use it to:
- resume a saved branch
- export or share a branch
- replay a branch timeline
- inspect previous work before continuing
Typical commands:
/recall/resume/export/share
Control Lane
Control is the operations surface inside the product.
Use it to:
- connect xAI
- inspect runtime readiness
- change permission mode
- inspect MCP and browser state
- inspect harness and bench status
Typical commands:
/connect/status/trust/harness/bench
Mission Panel
The Mission panel lives inside Work.
Use it to:
- create a mission
- start, pause, or resume a mission run
- inspect mission progress without leaving your main working lane
Useful commands:
/mission/mission create/mission start/mission pause/mission resume
Models, Profiles, And Reasoning In The TUI
Use:
/modelto pick the xAI model/profileto pick a routing profile/thinkor/thinkingwhen the current model exposes a user-facing reasoning selector
Important behavior:
- some xAI models manage reasoning internally
- multi-agent depth is controlled through research profiles
- if a model family uses reasoning/non-reasoning variants, the TUI exposes that as a toggle instead of a free-form effort control
Things That Are Present But Limited
/voiceexists, but voice input is unavailable in the xAI-only build- browser-backed research can be partially or fully unavailable depending on runtime state
- some advanced panels stay sparse until you actually create sessions, artifacts, runs, or memory
Use Troubleshooting when a lane looks empty and you are not sure whether that is normal.