Skip to content
Trending storyDeveloping

Infini-AI-Lab releases Hermit, an open-source harness adapter framework for AI agents

2 articles2 sourcessince Oct 26Last article 3h ago ·

Overview

AISummary of 2 articles

Infini-AI-Lab has released Hermit, an Apache-2.0 Python framework that trains reusable harness adapters while the agent's model and harness stay frozen, according to MarkTechPost.

The first release, v0.1.0, shipped on October 25, 2026, and installs with pip install hermit-harness. Developers can adapt Codex or Claude Code agents to new tasks without fine-tuning or forking the harness.

Earlier, a post from Dongxi NLP (东锡) on X announced the same framework and linked to the code on GitHub, with no independent evaluation reported at that time. MarkTechPost reports self-reported gains on OfficeQA, CL-Bench and ServeLearnBench, including 0.795 accuracy for the MWM adapter on CL-Bench database exploration versus 0.580 for the in-context baseline. These figures come from the project's own reporting and have not been independently verified.

Written by AI from the articles below · updated Oct 29, 5:51 AM ET

Check the sources:

Latest developmentOct 29, 4:24 AM ET
Infini-AI-Lab releases Hermit, an open-source harness adapter framework for AI agents

Article timeline

The articles in this story. Times are ET.

Oct 29
  1. MarkTechPostNews
    Infini-AI-Lab releases Hermit, an open-source harness adapter framework for AI agents

    AIInfini-AI-Lab has released Hermit, an open-source Python framework that trains an external harness adapter while the agent's model and harness stay frozen. Developers can adapt Codex or Claude Code agents to new tasks without fine-tuning or forking the harness. The article reports self-reported gains on OfficeQA, CL-Bench and ServeLearnBench, including 0.795 accuracy for the MWM adapter on CL-Bench database exploration versus 0.580 for the in-context baseline.

Oct 26
  1. Dongxi NLPX
    Hermit open-sources a framework for evolving harness adapters

    AIDongxi NLP shares Hermit, a framework for continual learning and recursive self-improvement through harness adapter evolution. The adapters are trained as reusable components without changing the underlying model or harness. Hermit supports multiple methods, and its code is available on GitHub.

Heat Trend

Current comparable heat
9
Peak comparable heat
10Oct 26, 2:00 PM ET
Change in past 24 hours
+210%

The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.