An agent harness · 8 MB · one binary
Run your own Vela.
Vela is a self-hosted agent harness: one static Go binary holding the agent loop, the tool belt, and persistent memory. Point it at any OpenAI-compatible model, give it a chat front end, and it runs — on a server, a laptop, or a credit-card-sized RISC-V board drawing two watts.
> make riscv64 · scp · edit one env file · done
Install Vela
The agent loop, tools, and memory in one binary. Swappable model, Discord as the shipped front end. Start here — everything else is optional.
02 / OPTIONAL PLUGINAdd Holodex
A deploy target, not a requirement. Without it Vela still researches, writes, and publishes repos — she just has nowhere to host a live demo.
03 / OPTIONAL PLUGINAdd the Foundation
A Rails 8.1 template so generated apps start from something real: Material Design 3, auth, commerce, and a validated boot-time configuration.