An app environment is the server that runs agents and apps. This page says what you are looking after and sends you to the page for each task. It holds no reference material itself.
If something is broken right now, go to troubleshooting. The first thing to check is always curl <server>/api/health.
What you are running
One process, one thread, one SQLite file. The server does four things:
It answers the signed API and a WebSocket.
It runs the agents' model turns and scripts.
It polls for the events that fire triggers.
It keeps all of its own state in a database on local disk, plus one directory per agent for that agent's memory.
It has no pages for a browser. People use it through the Seed desktop app and the web app.
Around it the server needs:
an HM server to read and write documents through. This is a site, a desktop app's local API, or a public gateway;
a reverse proxy for TLS, when people reach it over a network;
optionally, two helper containers for web search and for reading web pages with a browser;
optionally, hardware virtualization on the host, so agents can run code.
Model provider keys are not part of its configuration. Each account stores its own keys through the signed API; see model providers.
Reading order
Environments, to find the setup you have.
Configuration and deploying, to start the server.
Endpoints and health, to check it.
Troubleshooting, when something is wrong.
The other pages are reference. Read them when you need them.
Pages
Setting up:
Environments: the five setups and the settings each one uses.
Self-hosting an app environment: the decisions for running the image on your own server.
Configuration: every environment variable and flag.
Deploying: the image, its version metadata, the CI workflows, and rollouts.
Web research backends: the SearXNG and Crawl4AI containers.
Code execution backend: what the host needs for sandboxes, and how to check them.
Watching:
Endpoints and health: every path, the base URL shapes, and the health, version, timing and metrics endpoints.
Logs: log levels and the list of log lines.
Startup and shutdown: the start order, the background tasks, the stop order.
Run queue: how many runs execute at once, restarts, retries, and when the queue pauses.
Looking after the data:
Persistence: the database schema, table by table, and the files on disk.
Database retention: what keeps the file from growing forever, and how to shrink it.
Repairing:
Troubleshooting: starts from what you see.
Debugging runs: stuck sessions, parents that never resume, scripts that fail.
Diagnosing a stalled server: a server that is up but does not answer.
See also
Running Seed, the hub for everything an operator runs
App environment, the program map
How agents work, the system from the inside
Security, the trust boundaries an operator relies on
Development and testing the app environment, for people changing the code
Do you like what you are reading? Subscribe to receive updates.
Unsubscribe anytime