Install endpoints for orvex AI tooling. Each script is a single bash file you can curl … | bash or download + inspect first.
curl -fsSL orvex.sh/install/omnigent | bashInstalls the orvex build of orvexai/omnigent with uv, from the prebuilt wheels on a tagged release — no Node, no build, a few seconds. Needs gh logged in to the orvexai org, since the repo is private. Releases are cut from mcp, where the agent/MCP work lands; orvex is kept lean and lags it. The orvex line numbers its own releases, tagged orvex-vX.Y.Z — upstream's version names the line we rebased onto and moves on the next rebase, so the tag carries the real one. Installs the newest published release; a freshly cut one stays a draft until a human publishes it. Add -s -- --release <x.y.z> to pin one, or -s -- --ref <branch> to build from source instead.
curl -fsSL orvex.sh/install/houston | bashProvisions and keeps current the full Orvex agent toolchain for a project: BMAD foundation and modules, Claude plugins, shared MCP servers (including per-repo Serena code-intelligence daemons), Linear and Docmost tracking, and Beads issue memory. The command above only installs the tool, to ~/.local/bin/houston-installer — it fetches and SHA-256-verifies the newest published release of orvexai/houston-installer but runs nothing. Set the machine up with houston-installer install, then provision a project with houston-installer repo-init in it; the binary self-updates on every run after that. Needs only curl; since the repo is private, an anonymous 404 falls back to gh logged in to the orvexai org. Windows: run irm orvex.sh/install/houston.ps1 | iex instead (provisional — not yet run on a real Windows machine).