Arcee AI Open-Sources NAC: An Agent Harness for Long-Running Complex Tasks
code_star · x · 2026-08-14
Arcee AI released NAC (Apache 2.0), an open-source agent runtime designed specifically for managing long-running, complex engineering tasks.
Core Problem & Approach:
- Traditional agents suffer from "context rot" during long tasks, as exploration, tool outputs, and failed attempts accumulate in a single context window, degrading performance and losing early details.
- NAC decouples the temporary context needed for actions from the persistent state needed to continue the workstream, effectively bypassing context window limits.
Usage:
- It can be used as a highly opinionated harness for long-running tasks.
- It can also be exposed to coding tools like Codex or Claude Code as an abstraction layer to delegate difficult work. For instance, a single Codex session can manage 4-5 NAC sessions running models like GLM or Deepseek for cross-repo tasks.
Related event: Arcee Open Sources NAC Long-Running Agent Framework(14 posts)→
More from coding & agent
- DIRECT: retrieval-based next-question suggestions for conversational agents — burkov · 2026-10-16
- Claude Code and Codex now autocomplete the user's next action — tokenbender · 2026-10-16
- Hermes Agent ships official Reverse Engineer Anything plugin with one-command setup — Teknium · 2026-10-16
- MLExpert: 35B MoE Agent Trained to Autonomously Handle Machine Learning Engineering — Shangheng Du · 2026-10-16
- Cowork-Gym: Alibaba Synthesizes 64.9K Agent Work Environments, Boosting Qwen by 8.7 Points — Anni Zou · 2026-10-16
- MMSearch-OPD: Joint RL and On-Policy Distillation for Long-Horizon Multimodal Search Agents — PekingUniversity · 2026-10-16