Mixture of Agents in Hermes Agent: Multi-Model Architecture for Agent Workflows

ArticleNous ResearchResource

Nous Research has documented the Mixture of Agents (MoA) feature in Hermes Agent. Operating as a virtual model provider, MoA runs multiple reference advisor models in parallel to inform a central aggregator model that handles tool execution and user responses.

Key points

  1. MoA operates as a standard provider with selectable presets across CLI, TUI, Desktop GUI, and Dashboard interfaces.
  2. Advisory reference models receive conversation history without system prompts or tool schemas, reducing token spend and strict-provider rejections.
  3. Only the aggregator model executes tool calls and generates the final assistant output.
  4. Advisor cadence can be tuned via fanout modes: user_turn (default since July 2026), per_iteration, or every_n intervals.
  5. A privacy filter masks credentials, emails, and formatted phone numbers across UI traces and aggregator prompts.
  6. In HermesBench evaluations, an Opus 4.8 aggregator with a GPT-5.5 reference reached a score of 0.8202, outperforming standalone Opus 4.8 (0.7607).

Why it matters

MoA combines diverse perspectives from leading frontier models within an iterative agent loop while safeguarding prompt cache stability and controlling operational costs.

Analysis

By treating MoA as a regular model provider rather than an external orchestration toolset, Hermes Agent streamlines multi-model workflows. The design prioritises efficiency: prompt caching remains unbroken because reference advice is injected only at the prompt tail. Furthermore, features like reference token caps and the shift from per-iteration to per-user-turn execution address the latency and cost penalties typical of multi-agent patterns.

Found in

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Similar in content, found by the AI search.

  • Article:hermes-agent.nousresearch.com

    Hermes Agent Introduces Bot Mode for Persistent Multi-Agent Workflows

    Hermes Agent has added Bot Mode to its desktop app, transforming configuration profiles into named, specialized bots with isolated memory, skills, and model settings. Bots operate in persistent canonical chats, run recurring automated routines, and can collaborate across multiple machines in group chats and direct messages.

    AI & LLMs· Announcement

  • Link:Nous Research

    Hermes Bots Guide: Practitioner's Guide to Multi-Agent Systems by Nous Research

    Nous Research has published a guide covering Bot Mode in Hermes Agent. The feature enables users to run isolated agent profiles as named, persistent bots that execute recurring routines, expand their own skills, and interact across more than 20 messaging platforms.

    AI & LLMs· Resource

  • Repository:NousResearch/hermes-agent

    Hermes Agent: Self-Improving Open Source AI Agent by Nous Research

    Hermes Agent is a cross-platform, self-improving AI agent designed with an autonomous learning loop, cross-session memory, and integration into multiple messaging platforms and terminal interfaces.

    240.8kstarsPython

    AI & LLMs· Tool

  • Link:Nous Research

    Hermes Agent: Self-Improving Autonomous AI Agent by Nous Research

    Nous Research has documented Hermes Agent, an autonomous AI agent featuring a built-in learning loop. It develops and refines skills through operational experience, retains context across sessions, and operates across diverse environments from local machines to serverless setups.

    AI & LLMs· Resource

  • Link:Nous Research

    Hermes Agent: Open-Source AI Agent by Nous Research

    Nous Research has released Hermes Agent v0.21.0 under the MIT licence. The tool serves as an extensible AI agent available across multiple operating systems and messaging channels, featuring persistent memory and isolated execution sandboxes.

    AI & LLMs· Tool

  • Repository:NousResearch/Hermes-Bot-Mode

    Hermes Bot Mode: Multi-Agent Roster for Hermes Desktop

    Hermes Bot Mode extends the Hermes Desktop client with a roster of named AI agents featuring custom avatars, profiles, routines, and group chats. The standalone repository is now archived because the functionality has graduated into a built-in default plugin within Hermes Desktop.

    659starsJavaScript

    AI & LLMs· Tool

Let's talk about your project

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.