Run SignalScout yourself

The self-hosted build is the whole product. Nothing is removed to make the cloud worth paying for. You run it on your own server, and you bring the keys for the data providers and the AI model you want to use.

What you need

• A server with 1 GB of memory, or more. Intel or ARM: the image is published for both, so a Raspberry Pi or a cheap ARM server works.
• Docker with Compose. You do not need git, Node or a build.
• A domain name, if other people or other devices will reach it. It must sit behind HTTPS before it has a public address.
• At least one data provider key and one AI model key. One Reddit provider and one model are enough to start. Providers and keys lists them.

What runs

SignalScout is one Postgres database and one application container. The queue lives in Postgres too, so there is no Redis and no second database to back up.

• app serves the screens and the API on port 3000, and runs the worker that polls, filters, scores and notifies.
• migrate updates the database schema before app starts, then stops.
• postgres holds everything: your keys, monitors, matches and the job queue.

The worker can run in its own container instead. That is a setting, not a different image: see WORKER_IN_PROCESS in the configuration reference.

The path through these pages

1. Install — five commands, and the two secrets.
2. Providers and keys — which provider fetches which platform.
3. AI models — which model does which job.
4. HTTPS and the proxy — before the server has a public address.
5. Email and webhooks — how matches reach you.

Then keep it running: what it costs, backups and upgrades, and troubleshooting.