System Requirements
Genfeed runs as one application container plus a PostgreSQL container. The figures below are ones we have measured; where we have not measured something, the table says so rather than guessing.
Requirements
| Item | Supported floor | Recommended |
|---|---|---|
| Container runtime | Docker Engine 24+ with the Compose v2 plugin, or Docker Desktop | Same |
| Free disk | About 8 GB | 10 GB, plus room for generated media and backups |
| CPU | Not yet published | TBD |
| RAM | Not yet published | TBD |
| CPU architecture | linux/amd64 image (Docker Desktop can emulate it on Apple Silicon) | Same |
| Host ports | 3000 (web), 3010 (API), 3014 (MCP) free | Same |
| Node.js | Not needed for the release bundle. >=24 <25 for npx @genfeedai/create | Node.js 24 |
| Genfeed Cloud account | Not required | Not required |
Disk covers the application image, PostgreSQL, and the media you generate, so
it grows with use. Host ports are overridable with SELFHOSTED_APP_PORT,
SELFHOSTED_API_PORT, and SELFHOSTED_MCP_PORT; see
Configuration.
What we have measured
- A clean Debian 13 machine with 8 CPUs and 15 GiB of RAM ran the stack.
- Idle PostgreSQL used about 59 MiB.
These are measurements from one machine, not a minimum. We have not yet run smaller hosts, so no RAM or CPU minimum is promised. If you run Genfeed on smaller hardware, tell us how it went .
Operating system notes
- Linux: install Docker Engine with the Compose v2 plugin.
- Debian 13: install Debian’s
docker-composepackage, which provides Compose v2. With Docker CE, usedocker-compose-pluginfrom Docker’s apt repository instead. See Troubleshooting. - Docker Desktop: can be used instead of Docker Engine. On macOS, verify the
release bundle with
shasum -a 256 -cinstead ofsha256sum -c.
Model providers
Genfeed itself does not need a GPU. The Community image does not start GPU or local-model services, so generation uses the provider keys you add. See Provider API keys.