Replace feature tables with a focused bulleted list highlighting deterministic graphs, human-in-the-loop, multi-model routing, cloud sandboxes, SSH/preview, git checkpointing, retros, API, single binary, and open source. Merge the stylesheet example into the workflow example for one cohesive story. Remove TOC, node types table, supported models section, and outdated access warning. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> |
||
|---|---|---|
| .cargo | ||
| .claude | ||
| .config | ||
| .github | ||
| apps | ||
| arc/workflows | ||
| bin | ||
| docker | ||
| docs | ||
| lib | ||
| skills/develop-web-game | ||
| test | ||
| .env.example | ||
| .gitignore | ||
| AGENTS.md | ||
| arc.toml | ||
| bun.lock | ||
| Cargo.lock | ||
| Cargo.toml | ||
| CLAUDE.md | ||
| LICENSE.md | ||
| package.json | ||
| README.md | ||
The software factory for small teams of expert engineers
AI coding agents are powerful but unpredictable. You either babysit every step or review a 50-file diff you don't trust. Arc gives you a middle path: define the process as a graph, let agents execute it, and intervene only where it matters.
Key Features
- Deterministic workflow graphs -- Define pipelines in Graphviz DOT with branching, loops, parallelism, and human gates. Diffable, reviewable, version-controlled.
- Human-in-the-loop -- Approval gates pause for human decisions. Steer running agents mid-turn. Interview steps collect structured input. You stay in control without babysitting.
- Multi-model routing -- Route each node to the right model and provider with CSS-like stylesheets. Cheap models for boilerplate, frontier models for hard reasoning, with automatic provider fallback.
- Cloud sandboxes -- Run agents in isolated Daytona cloud VMs with snapshot-based setup, network controls, and automatic cleanup.
- SSH access and preview links -- Shell into running sandboxes with
arc sshand expose ports witharc previewfor live debugging while workflows run. - Git checkpointing -- Every stage commits to a branch. Inspect changes, revert mistakes, or resume interrupted runs exactly where they left off.
- Automatic retros -- Each run generates a retrospective with cost, duration, files touched, and an LLM-written narrative rating smoothness and flagging friction points.
- Comprehensive API --
arc serveexposes a full REST API with SSE event streaming and a React web UI. Run workflows programmatically, build integrations, or operate Arc as a service. - Single binary, no runtime -- One compiled Rust executable with zero dependencies. No Python, no Node, no Docker required to get started.
- Open source (MIT) -- Full source code, no vendor lock-in. Self-host, fork, or extend to fit your workflow.
Read the full documentation for details.
Quick Start
Install
Download the latest release for your platform:
# macOS (Apple Silicon)
curl -fsSL https://github.com/brynary/arc/releases/latest/download/arc-aarch64-apple-darwin.tar.gz | tar xz
sudo mv arc /usr/local/bin/
# Linux (x86_64)
curl -fsSL https://github.com/brynary/arc/releases/latest/download/arc-x86_64-unknown-linux-gnu.tar.gz | tar xz
sudo mv arc /usr/local/bin/
Or download directly from GitHub Releases.
Build from source
Requires Rust (latest stable):
git clone https://github.com/brynary/arc.git
cd arc
cargo build --release
# Binary is at ./target/release/arc
Run the setup wizard
arc install
Initialize your project
cd my-repo/
arc init
Run your first workflow
arc run hello
Example Workflow
A plan-approve-implement workflow where a human reviews the plan before the agent writes code:
digraph PlanImplement {
graph [
goal="Plan, approve, implement, and simplify a change"
model_stylesheet="
* { llm_model: claude-haiku-4-5; reasoning_effort: low; }
.coding { llm_model: claude-sonnet-4-5; reasoning_effort: high; }
"
]
start [shape=Mdiamond, label="Start"]
exit [shape=Msquare, label="Exit"]
plan [label="Plan", prompt="Analyze the goal and codebase. Write a step-by-step plan.", reasoning_effort="high"]
approve [shape=hexagon, label="Approve Plan"]
implement [label="Implement", class="coding", prompt="Read plan.md and implement every step."]
simplify [label="Simplify", class="coding", prompt="Review the changes for clarity and correctness."]
start -> plan -> approve
approve -> implement [label="[A] Approve"]
approve -> plan [label="[R] Revise"]
implement -> simplify -> exit
}
Agents run as multi-turn LLM sessions with tool access. Human gates (hexagon) pause for approval. The stylesheet routes planning to a cheap model and coding to a frontier model. See the DOT language reference for the full syntax.
System Requirements
- macOS (Apple Silicon) or Linux (x86_64)
- At least one LLM provider API key (Anthropic, OpenAI, or Gemini)
- Git (for checkpoint and resume)
- Rust (only if building from source)
Help or Feedback
- Read the documentation
- Bug reports via GitHub Issues
- Feature requests via GitHub Issues
- Email bryan@qlty.sh for questions
Contributing
See CLAUDE.md for build commands, architecture overview, and development conventions.
License
Arc is licensed under the MIT License.