The OrchestratorOverviewHow It Connects
The StudioOverviewExtensionsHow It Works
ResourcesBlogFAQAbout
Get in touch
Early Access
Studio Extension

Terminal

Command detection, process management, agent-callable tools

Your agents can run your project's build, test, and deploy commands directly — without guessing syntax or relying on cloud CI. The Terminal extension detects project commands from your config files and exposes them as structured tools agents can call.

Command list and timeline
Command list and timeline
Terminal tools in chat
Terminal tools in chat
Manual terminal session
Manual terminal session

All detected project commands in one view — with execution history and timeline tracking.

Capabilities

What it does

Command Detection

Scans package.json scripts, Makefile targets, and docker-compose services. Agents see a structured list of available commands — no guessing, no hallucinated syntax.

Process Lifecycle

Start, monitor, and stop processes with full lifecycle tracking. Configurable shell, working directory, and output buffer. Agents can check if a process is running before starting another.

Interactive Shells

Full interactive terminal sessions via xterm.js and node-pty. Agents can run commands that require interaction — prompts, confirmations, multi-step workflows.

Streaming Progress

Command output streams over HTTP as NDJSON — progress updates, warnings, user prompts, and final results. Agents react to output in real time, not after completion.

How it works

From install to first use.

1
Detect project commandsOn workspace open, the plugin scans package.json, Makefile, and docker-compose.yml. It builds a structured catalog of available commands with descriptions.
2
Agents call tools, not shell commandsInstead of generating raw shell commands (which can be wrong or dangerous), agents call structured tools — run_tests, start_server, build_project — with validated parameters.
3
Stream results backOutput streams as NDJSON over HTTP. Agents see progress in real time and can react — retry on failure, parse test results, or escalate errors.
4
Manage process lifecycleLong-running processes (dev servers, watchers) are tracked. Agents can check status, read recent output, or stop processes cleanly.
Why local matters

Run locally, not in the cloud.

Your build, test, and deploy commands run on your machine — not in a cloud CI pipeline. Faster feedback, no network latency, no secrets exposed to remote runners.

LocalAll commands execute on your machine
0secrets sent to cloud runners
<1slatency vs cloud CI roundtrips
Real-timestreaming output, not batch results

Terminal ships with the Studio. No extra install, no extra cost.

Get early access All extensions