@gobing-ai/spur

@gobing-ai/spur

The spur command — a local-first harness for mainstream coding agents (Claude Code, Codex, Gemini CLI, Antigravity, pi, OpenCode, OpenClaw). Spur is not a coding agent; it wraps the agents you already have with constraint checking, workflow orchestration, agent health checks, and conversation-history analytics.

Install

Pick the path that matches your setup. Both give you a global spur command and seed the default config into ~/.config/spur/ on first run.

If you have Bun (bun >= 1.3.0) — recommended

The npm package ships a Bun bundle and runs under the Bun runtime you already have.

# Install globally → use the `spur` command everywhere
bun install -g @gobing-ai/spur
spur --help

# …or run ad-hoc with no install
bunx @gobing-ai/spur --help

If you don't have Bun — standalone binary

A self-contained executable (Bun embedded) for macOS and Linux. No runtime to install.

# One-liner: downloads the binary, puts it on PATH, seeds config
curl -fsSL https://raw.githubusercontent.com/gobing-ai/spur/main/scripts/install.sh | sh
spur --help

The installer drops spur into ~/.local/bin by default. Override the target with SPUR_INSTALL, or pin a release with SPUR_VERSION:

SPUR_INSTALL=/usr/local/bin SPUR_VERSION='@gobing-ai/spur-v0.1.7' \
  sh -c "$(curl -fsSL https://raw.githubusercontent.com/gobing-ai/spur/main/scripts/install.sh)"

If ~/.local/bin isn't already on your PATH, add it (e.g. export PATH="$HOME/.local/bin:$PATH" in your shell profile). Supported targets: darwin-arm64, darwin-x64, linux-arm64, linux-x64. Windows: use the Bun path under WSL.

First run

spur init     # seeds ~/.config/spur/ (config.yaml, rules/, workflows/) — idempotent, never clobbers

The standalone installer runs spur init for you; the Bun install does it on your first spur init.

Usage

spur init                                  # scaffold .spur/ + seed global rules
spur rule run --preset recommended         # evaluate constraint rules
spur workflow run <workflow.yaml>          # run an FSM workflow
spur agent run "<prompt>" --agent auto     # execute a prompt via a coding agent
spur agent doctor                          # check agent readiness
spur history import --source claude --root <path>
spur history analyze
spur status

Every command supports --json for machine-readable output.

Documentation

Full docs, architecture, and the complete command surface live in the Spur repository.

License

Apache-2.0 Robin Min