Installation And Launch#
Agent Master ships with pyCSAMT and uses the same optional application extra as the desktop, web, and MapView apps — one install covers every application surface.
Install The App Extra#
pip install "pycsamt[app]"
For development from a source checkout:
pip install -e ".[app,dev]"
The extra brings the Dash/Plotly stack the app renders with; the scientific core and the agents are part of the base install. To actually drive the LLM-backed features you also need a model provider and API key — see LLM Configuration.
Launch#
pycsamt-agent
The launcher starts a local server on http://127.0.0.1:8765 and opens the
app in your default browser. The module form launches the same app:
python -m pycsamt.app.agent_master
Leave the terminal open while you work — it hosts the server. Press Ctrl+C
to stop.
Options#
Option |
Effect |
|---|---|
|
Bind address (default |
|
HTTP port (default |
|
Start the server without opening a browser tab. |
|
Enable Dash debug / hot-reload — useful only when developing the app itself. |
Examples:
pycsamt-agent --port 8770 # serve on a different port
pycsamt-agent --no-browser # headless / remote host
pycsamt-agent --host 0.0.0.0 # allow LAN access (trusted networks only)
The Welcome Screen#
On first load, Agent Master shows a welcome screen with a single call to action.
The welcome screen summarises the four things Agent Master does — Load EDI, Chat & Plan, AI Inversion, and Reports — and starts the session with Start Agent Master.#
Verify#
After launch, confirm that:
the terminal reports the server is running on
http://127.0.0.1:8765;the browser opens to the welcome screen (or copy the URL manually on a headless host);
Settings (the gear icon) opens and shows the LLM provider options.
If the browser does not open, the page is blank, or the port is busy, see Troubleshooting.
Next Steps#
LLM Configuration – connect Claude, OpenAI, Gemini, or DeepSeek.
Welcome Screen And Chat – the interface, loading data, and chatting.