You configure the app environment with environment variables and command-line flags. This page lists all of them. For example, SEED_APPS_HTTP_PORT=3051 makes the server listen on port 3051, and so does the flag --server-port 3051.
The first thing to check: /api/health reports several of the values in use, so you can see what a running server was started with. The code reads every setting in one file, apps/src/config.ts. Environments says which settings each setup uses, and operations sends you to the task pages.
Names
Each setting is an environment variable named SEED_APPS_<NAME>. The server also reads the older spelling SEED_AGENTS_<NAME> for every setting. When both are set, the SEED_APPS_ name wins. The self-hosting compose files and the hosted Seed servers set the older spelling. These pages write the SEED_APPS_ form.
Environment variables
Environment variable | Default | Purpose |
|---|---|---|
|
| HTTP bind hostname. |
|
| HTTP port. |
|
| SQLite DB path. |
|
| Data directory. |
|
| The Seed API the server reads and writes through: the |
| (unset) | The service secret of a members-only site. It is sent as |
| (unset) | Where clients reach this server, such as |
| HM server URL | The origin for direct |
| the Agent Guide on hyper.media | The document a new agent's system prompt embeds, the Agent Guide. Reported as |
| (unset) | The account of the knowledge base space. Reads of |
| (unset) | A file that names the knowledge base account, read on every use. Dev points it at the file |
|
| How often the activity monitor and the schedule monitor look for work. These are the tasks that fire triggers. |
|
| ActivityFeed page size. |
|
| Max pages fetched per poll. |
| (unset) | Self-hosted SearXNG base URL. It makes |
| (unset) | Self-hosted Crawl4AI base URL. With it, a web read can fall back to a real browser for pages that need one. |
| (unset) | Bearer token for Crawl4AI (required by Crawl4AI >= 0.9). |
| (off) | Offer provider OAuth sign-in ("Sign in with ChatGPT"). Only |
|
| Name untitled sessions with a dedicated model call. Set exactly |
|
| Code-execution backend for |
|
| OCI image used for execution sandboxes. |
|
| OCI image for the |
|
| Virtual CPUs per execution sandbox. |
|
| Guest memory per execution sandbox (MiB). |
|
| Default per-execution timeout (tool may request up to 300s). |
|
| Sandbox internet access. Set |
|
| Comma-separated DNS resolvers used inside execution sandboxes. |
| CPU count minus 2, at least 1 | Host-wide cap on sandboxes in use at once; a call past the cap waits for a slot. |
|
| How long a call waits for a sandbox slot under the host-wide cap before failing with a retryable error. |
|
| How long a call waits for the agent's memory to be free of other sandboxes (another session, a delegated child, a trigger firing of the same agent) before failing with a retryable error. Long by default so those calls queue behind the agent's own running execution instead of failing. |
| off | Set |
|
| Warm pool only: pooled VMs retained; extra concurrent calls boot transient VMs rather than wait. |
|
| Warm pool only: how long a parked VM may sit idle before it is disposed. |
|
| Warm pool only: maximum age of a pooled VM, enforced between calls. |
|
| How often the exec canary boots a fresh sandbox and runs one command, to prove code execution still works. The hosted production and staging servers set |
|
| Free space the database's file system must have before the queue starts another run; below it runs stay queued and |
|
| Agent runs that hold a provider stream at once. See the run queue. |
|
| Script runs executed at once. They hold no provider stream. |
| off | Set |
|
| SQLite page cache (private memory, counted against the container limit). See persistence. |
|
| Memory-mapped reads (reclaimable page cache), capped by SQLite's compiled maximum; the value in effect is logged at startup. |
|
| Size the WAL file is truncated back to after a checkpoint. |
|
| The retention task: |
|
| A session idle this long loses its durable store, which its next turn rebuilds from the log. |
|
| A workflow run finished this long ago has the large payloads in its journal cut to previews. |
|
| Disk budget for retention work. |
|
| Durable stores larger than this are left for the offline pass (a store is dropped in one transaction). |
|
| Time between retention passes; the first runs 15 minutes after start. |
|
| Longest string (in thousands of characters) a stored tool output keeps; longer ones keep their head and a marker. The model never reads more than 32 KiB of a result either way. |
|
| Minimum log level ( |
Five more variables are set by the build, not by an operator. The image build writes SEED_APPS_VERSION, SEED_APPS_COMMIT, SEED_APPS_BRANCH and SEED_APPS_DATE, which the version endpoint reports (see deploying). The desktop build puts SEED_APPS_STANDALONE into the bundled binary so that it starts as a server.
Where a row says what the hosted Seed servers set, it describes how those servers are run. That can change, and this repository cannot confirm it.
Command-line flags
A flag wins over the environment variable, and both win over the default:
bun src/main.ts \
--server-hostname 127.0.0.1 \
--server-port 3050 \
--db-path ./data/agents.sqlite \
--data-dir ./data \
--hm-server-url https://hyper.media \
--ipfs-server-url https://hyper.mediaEvery environment variable in the table has a matching flag, unless its row says it has none. Examples are --searxng-url, --exec-backend, --exec-ts-image, --subscription-auth, --session-title-generation and --max-concurrent-model-runs. The full list is the Flags type in config.ts.
An unknown flag, or a flag with an empty value, stops the server at startup. So a typing mistake fails at once, and the server never runs with a default you did not choose.
See also
Operations, the overview of running an app environment
Environments, the settings each deployment turns on
Endpoints and health, where the server reports the values in effect
Web research backends and the code execution backend, the two optional subsystems
Database retention, what the retention settings remove
App environment, the program map
Do you like what you are reading? Subscribe to receive updates.
Unsubscribe anytime