nac
Give AI agents ambitious work without losing the plot. nac is an open-source harness for long-running tasks, using a central orchestrator, threads, and structur
What is nac?
Give AI agents ambitious work without losing the plot. nac is an open-source harness for long-running tasks, using a central orchestrator, threads, and structured episodes to stay aligned with your intent.
Overview of nac
What it is
(nac is an open-source agent harness for longer, ambitious tasks such as experiments, training runs, infrastructure, and prototyping. It uses a thread-and-episode architecture: a central orchestrator plans and decomposes work but cannot execute commands or edit files; it only launches threads, which return episodes — structured summaries of what they accomplished. The project takes inspiration from slate, nanocode, and pi.)
Key features
- Thread-and-episode architecture with a central orchestrator that plans work and launches threads, receiving structured episode summaries.
- Open-source (Apache 2.0) and self-hostable via a local dashboard.
- Supports multiple model providers: Arcee, ChatGPT Codex, and any OpenAI-compatible API (DeepSeek, Fireworks, Together, OpenAI, Anthropic, etc.).
- MCP integration for connecting with agent harnesses and tools.
- Skills and sandbox system for extending functionality and safe execution.
- Native upgrade command (
nac-web upgrade) for updating to the latest stable release.
Use cases
Long-running experiments, training runs, infrastructure management, and prototyping that must stay aligned with the original intent. Suitable for AI agents that need to execute ambitious tasks over extended periods while maintaining focus.
Getting started
Install the latest stable release: sh curl -fsSL https://raw.githubusercontent.com/arcee-ai/nac/main/scripts/install.sh | sh The installer places nac-web in $HOME/.local/bin. Add that directory to your PATH if needed, then start the dashboard from your project: sh nac-web Navigate to the interface in your browser (default: http://127.0.0.1:3210).
License and pricing
Licensed under Apache 2.0. The project is open source and free to use.
Based on the project's public README and repository metadata.