--- title: Infini-AI-Lab Releases Hermit, an Open-Source Harness... description: Infini-AI-Lab released Hermit, an open-source framework that trains an external 'harness adapter' while an AI agent's model and harness stay frozen. Version... canonical: https://daily.dev/posts/infini-ai-lab-releases-hermit-an-open-source-harness-adapter-framework-for-ai-agents-1btxpfdmx twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Infini-AI-Lab Releases Hermit, an Open-Source Harness Adapter Framework for AI Agents | daily.dev og:description: Infini-AI-Lab released Hermit, an open-source framework that trains an external 'harness adapter' while an AI agent's model and harness stay frozen. Version... og:url: https://daily.dev/posts/infini-ai-lab-releases-hermit-an-open-source-harness-adapter-framework-for-ai-agents-1btxpfdmx og:image: https://api.daily.dev/og/posts/1BtxPFDmx.png og:image:alt: Infini-AI-Lab Releases Hermit, an Open-Source Harness Adapter Framework for AI Agents og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Infini-AI-Lab Releases Hermit, an Open-Source Harness Adapter Framework for AI Agents **[Machine Learning News](https://daily.dev/sources/mlnews)** · 5 min read · 0 upvotes · 0 comments ## Summary Infini-AI-Lab released Hermit, an open-source framework that trains an external 'harness adapter' while an AI agent's model and harness stay frozen. Version 0.1.0 shipped October 25, 2026, installable via pip install hermit-harness under Apache-2.0. It adapts agents like Codex or Claude Code to new tasks without fine-tuning or forking the harness. Five evolution methods ship, including GEPA, SkillOpt, MWM, Harness² and Continual Harness. On CL-Bench database exploration, the MWM adapter hits 0.795 accuracy versus a 0.580 in-context baseline. Results are mostly self-reported single runs using GLM 5.3 Flash. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is a harness adapter in AI agent frameworks like Hermit? A harness adapter is a trainable external component that adapts an existing agent without modifying its base model or harness. It can hold instructions, skills, memory, tools, executable code, and agent coordination logic, even an entire repository. Unlike a model adapter such as LoRA, which changes the model, a harness adapter wraps around a frozen model and frozen harness, specifying which parts are learnable. _Developers customizing Codex or Claude Code agents can track emerging adapter techniques like this on daily.dev._ ### How does Hermit's MWM adapter perform on CL-Bench compared to in-context learning? The MetaWorld Model (MWM) adapter reaches 0.795 accuracy on CL-Bench database exploration, compared to 0.580 for the in-context learning baseline. On exploitable poker, MWM wins 3.08 big blinds per hand versus 2.51 for in-context learning. These results use GLM 5.3 Flash as the acting model and are self-reported, mostly from single runs. _Teams evaluating agent optimization methods can compare benchmark claims like these on daily.dev._ ### Which AI coding agents can Hermit deploy trained adapters to? Hermit's deploy command writes a workspace for Codex, Claude Code, or Pi, and its convert command exports a trained adapter as a native plugin for Codex, Claude Code, Pi, or Hermes. A finished example adapter coordinates five Codex agents collaborating to prove a single Lean theorem together. _Engineers wiring adapters into Codex or Claude Code workflows can follow framework updates like this on daily.dev._ ## Similar posts on daily.dev - [Harness Engineering](https://daily.dev/posts/harness-engineering-l1sahcwor) · Martin Fowler · 4 upvotes · 0 comments - [Hermes Agent Masterclass](https://daily.dev/posts/hermes-agent-masterclass-k73au5ews) · Daily Dose of Data Science \| Avi Chawla \| Substack · 4 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#ai-agents](https://daily.dev/tags/ai-agents), [#claude-code](https://daily.dev/tags/claude-code), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/infini-ai-lab-releases-hermit-an-open-source-harness-adapter-framework-for-ai-agents-1btxpfdmx) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Infini-AI-Lab Releases Hermit, an Open-Source Harness Adapter Framework for AI Agents","url":"https://daily.dev/posts/infini-ai-lab-releases-hermit-an-open-source-harness-adapter-framework-for-ai-agents-1btxpfdmx","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/infini-ai-lab-releases-hermit-an-open-source-harness-adapter-framework-for-ai-agents-1btxpfdmx"},"datePublished":"2026-10-29T08:29:45.934Z","dateModified":"2026-10-29T08:30:31.873Z","description":"Infini-AI-Lab released Hermit, an open-source framework that trains an external 'harness adapter' while an AI agent's model and harness stay frozen. Version...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f625c5e72172cf66c8b2847db917d106?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f625c5e72172cf66c8b2847db917d106?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Machine Learning News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Machine Learning News","logo":"https://media.daily.dev/image/upload/s--0PoEtdkd--/f_auto/v1750409401/squads/marktechpost?_a=BAMClqZW0","url":"https://daily.dev/squads/mlnews"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/infini-ai-lab-releases-hermit-an-open-source-harness-adapter-framework-for-ai-agents-1btxpfdmx","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,ai-agents,claude-code,reinforcement-learning","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Machine Learning News","item":"https://daily.dev/squads/mlnews"},{"@type":"ListItem","position":3,"name":"Infini-AI-Lab Releases Hermit, an Open-Source Harness Adapter Framework for AI Agents"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/infini-ai-lab-releases-hermit-an-open-source-harness-adapter-framework-for-ai-agents-1btxpfdmx#faq","mainEntity":[{"@type":"Question","name":"What is a harness adapter in AI agent frameworks like Hermit?","acceptedAnswer":{"@type":"Answer","text":"A harness adapter is a trainable external component that adapts an existing agent without modifying its base model or harness. It can hold instructions, skills, memory, tools, executable code, and agent coordination logic, even an entire repository. Unlike a model adapter such as LoRA, which changes the model, a harness adapter wraps around a frozen model and frozen harness, specifying which parts are learnable. Developers customizing Codex or Claude Code agents can track emerging adapter techniques like this on daily.dev."}},{"@type":"Question","name":"How does Hermit's MWM adapter perform on CL-Bench compared to in-context learning?","acceptedAnswer":{"@type":"Answer","text":"The MetaWorld Model (MWM) adapter reaches 0.795 accuracy on CL-Bench database exploration, compared to 0.580 for the in-context learning baseline. On exploitable poker, MWM wins 3.08 big blinds per hand versus 2.51 for in-context learning. These results use GLM 5.3 Flash as the acting model and are self-reported, mostly from single runs. Teams evaluating agent optimization methods can compare benchmark claims like these on daily.dev."}},{"@type":"Question","name":"Which AI coding agents can Hermit deploy trained adapters to?","acceptedAnswer":{"@type":"Answer","text":"Hermit's deploy command writes a workspace for Codex, Claude Code, or Pi, and its convert command exports a trained adapter as a native plugin for Codex, Claude Code, Pi, or Hermes. A finished example adapter coordinates five Codex agents collaborating to prove a single Lean theorem together. Engineers wiring adapters into Codex or Claude Code workflows can follow framework updates like this on daily.dev."}}]} ```