System & Control APIs
Endpoints for metrics, screen vision, computer control, LLM access, settings, and telemetry.
Beyond chat and widgets, the agent exposes endpoints for system metrics, screen analysis, agentic computer control, direct LLM access, and configuration.
System metrics
GET /api/system/metricsendpointLive CPU, RAM, and disk metrics. Backs the system dashboard and the df / free
fast-path commands.
curl http://127.0.0.1:8000/api/system/metricsScreen vision
POST /api/screen/analyzeendpointCapture the active screen and analyze it with the vision model — used for "what's on my screen?" style queries and the screen-reader plugin.
POST /api/screen/insertendpointInsert/paste generated text into the currently focused application.
Computer control
The agentic control loop that powers Computer Control. The agent takes a screenshot, decides the next GUI action, executes it, and repeats.
POST /api/computer/executeendpointStart an agentic computer-control task (open apps, click, type, run terminal commands, drive an IDE).
POST /api/computer/stopendpointStop the running control loop.
Computer control can click, type, and run terminal commands on your machine. Treat it like handing over the keyboard — review what you ask it to do.
POST /api/overlay/showendpointShow a transient overlay UI on the desktop.
POST /project/setupendpointScaffold a new project (backed by the project_setup plugin).
Direct LLM & text tools
POST /api/llmendpointMake a direct LLM call through the agent's configured provider — handy for scripting
against whatever model LLM_MODEL points at.
POST /api/assistendpointGeneral text assistance (used by the writing assistant widget).
POST /api/transformendpointTransform selected text — improve, shorten, expand, fix grammar, continue.
Files & ingestion
POST /api/uploadendpointUpload a file to the agent.
POST /api/file-ingestendpointIngest a file's contents into context (parsed via the document handlers).
Settings
GET /api/settingsendpointRead current settings.
POST /api/settingsendpointUpdate settings. Because LLM config values are live properties, updating provider/key/ model takes effect without restarting the agent.
curl -X POST http://127.0.0.1:8000/api/settings \
-H "Content-Type: application/json" \
-d '{"LLM_MODEL": "anthropic/claude-sonnet-4-6"}'Telemetry & notifications
POST /api/ui_telemetryendpointThe shell reports render successes/failures here so bad component shapes can be found.
GET /api/ui_telemetryendpointRead accumulated render telemetry.
GET /api/notificationsendpointNotifications raised by the proactive monitor (calendar reminders, system-health alerts). See.
GET /api/dashboardendpointAggregated dashboard data.
Inbound webhooks
POST /webhook/google-chatendpointReceive Google Chat events.
POST /webhook/teamsendpointReceive Microsoft Teams events.
POST /cli/dispatchendpointDispatch a command from the bundled megha CLI.