grid shell is the entry point to the platform — install it with the
one-liner from the portal, run grid, and type help at any time. This
page lists every available command.
Robots
| Command | Description |
|---|---|
robot add [--config <yaml>] [--local [--root <dir> --dry-run]] [--existing-config overwrite] [--take-over] | Onboard a robot — the wizard in the REPL; headless calls require —config; SSH needs a connection section; —local is dev-profile-only; —root <dir> —dry-run (paired) rehearses it under a scratch root; —take-over answers the take-over confirm headlessly when the robot is already set up and its edge service is running |
robot configure <robot> [--config <yaml>] | Re-run install → enable → verify over an explicit SSH connection, or apply a full profile and its connection with —config (confirms; headless calls require —config) |
robot list | Connected robots — name, owner, team or personal, when its config last changed, and type |
robot show <robot> | One connected robot in detail: type, args, sensors and capabilities from its registration |
robot status <robot> [--config <yaml>] | Inspect service state, config status and versions over an explicit SSH connection; the REPL prompts, and headless calls require —config |
robot logs <robot> [--config <yaml>] [--follow] | Read the edge service’s journal over an explicit SSH connection — the last 200 lines, or follow until Esc; the REPL prompts, and headless calls require —config |
robot stop <robot> | Stop a robot you hold and release it — closes this machine’s session first |
Skills
| Command | Description |
|---|---|
skill run [path] [--robot <robot>] [--dev | --deploy] [-- script-args...] | Choose an online robot and run a Python file; —dev opens a workspace (path optional, defaults to the current folder); —deploy confirms. Headless runs require —robot |
skill stop [run-id] | Stop a run or development session started from this machine — confirms for deployments |
skill logs [run-id] [--follow] | Logs of a run — the latest one when no id is given |
Simulation
| Command | Description |
|---|---|
sim start [--local] [--name|-n <name>] [--resume <id>] [--config <yaml|json>] | Allocate a simulator runtime — robot → simulator → scene wizard, or a config file; —resume restores a stopped one |
sim list | Every sim runtime you can reach — name, simulator, state, age |
sim open [id] | Open a running sim’s browser workspace; with one running, the id is optional |
sim stop [id] | Release a sim runtime and its GPU — confirms; -y to skip |
Navigation
| Command | Description |
|---|---|
nav start [--no-open] | Start the Navigation editor in the shared local web app using the active cluster login. Keep GRID running while editing |
nav view <robot> [--no-open] | Visualize the live position, route and mission of the robot in this GRID process’s skill run —dev session, in the shared local web app. Run it in the REPL that holds the session; closing the page leaves the robot running |
Environment
| Command | Description |
|---|---|
context list | Every known context — host and active marker; kind (cluster | node | local) in dev builds |
context show <host> | One context in detail: origin, portal, Cortex and signaling endpoints, auth state |
context use [host] | Switch the active context; with no host, open the REPL picker; headless calls require a host |
context add --url <portal-url> | Register a cluster by its portal URL and switch to it; re-adding a known one just switches |
context remove <host> | Forget a cluster — its session and API key are cleared too |
key show | The public prefix of the GRID API key stored for the active cluster |
key remove | Forget the active cluster’s GRID API key so the startup gate asks for a new one; the session is kept |
login [--headless [--url <portal-url>] [--timeout <s>]] | Log in to the active GRID cluster in your browser; —headless prints the authorize URL and waits |
logout | Drop the active cluster’s session locally — your API key and context are kept |
status | Where you are and what is running from this machine — active context, robot session, skill runs and recorded simulations with live state |
Arms
| Command | Description |
|---|---|
device add | Register a leader device on this machine |
device list | List the leader devices registered on this machine |
device calibrate [device] [--robot <robot>] | Align a leader device against a robot, once per physical setup |
device check [robot] [--leader <device>] | Report whether teleop would start — the robot is never moved |
device teleop [robot] [--leader <device>] [--dataset <name>] [--task <text>] [--max-image-size <px>] | Teleoperate a robot from a registered leader device |
dataset list | List datasets recorded on this machine |
dataset visualize <dataset> [episode] [--no-open] | Play a recorded episode back — cameras and joint plots on one timeline |
finetune submit <dataset> [--name <label>] [--robot-type <type>] [--steps <n>] [--batch-size <n>] [--action-horizon <n>] [--save-after-step <n>] [--save-interval <n>] | Start a training run, uploading the dataset first if the cloud lacks it; the REPL opens a form for the knobs |
finetune list | List finetune jobs submitted from this account |
finetune cancel <job-id> | Stop a running finetune job |
Troubleshooting
| Command | Description |
|---|---|
error | Show details of the last error |
diagnostics | Bundle logs + config into an unscoped tarball for support; takes no arguments |
debug [on|off] | Toggle verbose logging with inline stack traces |
Utilities
| Command | Description |
|---|---|
help [command] | Show help information |
theme [grid|mono] | Switch the color theme for this session — mono renders grayscale |
clear | Clear the terminal |
exit | Exit GRID CLI |