Hermes Atlas
Multi-agent & orchestration · works with Hermes Agent

Beastmode

lac5q/beastmode

Mixture-of-Agents orchestration framework splitting design and execution across frontier and cheap models

In short

Beastmode is a Mixture-of-Agents (MofA) orchestration framework for AI-assisted software development that runs with Hermes Agent, OpenClaw, Claude Code, Codex and others. Frontier models design and review while cheaper executor models implement and verify.

What Beastmode does

Beastmode is a Mixture-of-Agents orchestration framework for AI-assisted software development. It divides work like a kitchen: frontier models act as chefs who design, taste-test and sign off, while cheaper executor models act as line cooks who implement and run verification. It routes implementation and mechanical validation to an economy lane (gpt-6-luna) and reserves frontier models such as Claude Fable, Kimi 3, Opus and Codex for design, judgment and review.

It runs with Hermes Agent, OpenClaw, Claude Code, Codex, Qwen, LangGraph and pi-coding-agent, and talks to editors over ACP. The README lists mandatory acceptance contracts, adversarial review and merge gates, plus a self-improvement loop that records lessons and promotes winning patterns into skills or config. A bm launcher handles install, upgrade, status, doctor, version and uninstall, using immutable versioned snapshots behind a current symlink tracked by a manifest, so upgrades are atomic and rollback is a symlink fix. A bm command takes a task, prints an ETA and can run on a chosen harness, for example --harness hermes for async parallel sub-agents.

Key features

  • Chef and line-cook split between frontier design models and cheaper executor models
  • Acceptance contracts, adversarial review and merge gates
  • Self-improvement loop that promotes recurring patterns into skills or config
  • Harness choice with --harness, including hermes, claude, codex and langgraph
  • Installer with immutable versioned snapshots, atomic upgrades and a manifest-driven uninstall
  • bm status and bm doctor commands for checking the install

When to use it

  • Cutting model cost by sending implementation to a cheaper lane while keeping review on frontier models
  • Running parallel Hermes sub-agents for a development task
  • Rolling back to an earlier framework snapshot after a bad upgrade

Who it is for: Developers who orchestrate several coding agents and want cost routing, review gates and versioned installs.

How it fits with Hermes Agent

It supports Hermes Agent as one of several harnesses, with a --harness hermes option for async parallel sub-agents.

How to install Beastmode

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

./scripts/install-beastmode.sh
./scripts/install-beastmode.sh doctor

FAQ

What is Beastmode?

Beastmode is a Mixture-of-Agents orchestration framework that splits software work between frontier models for design and review and cheaper models for execution. It adds acceptance contracts and merge gates.

Does Beastmode work with Hermes Agent?

Yes, the README says it runs with Hermes Agent as well as OpenClaw, Claude Code, Codex, Qwen, LangGraph and pi-coding-agent. The --harness hermes option runs async parallel sub-agents.

How do I install Beastmode?

From a checkout, run ./scripts/install-beastmode.sh, which installs the bm launcher and runtime to ~/.local. The doctor command then checks prerequisites and snapshot integrity.

Similar multi-agent for Hermes Agent

All multi-agent

Related guides: Build agent teams with the Hermes kanban board · Run multiple Hermes agents with profiles · Connect Hermes agents on several machines with Hermes Desktop