Agent Apprenticeship: research project for Hermes Agent

Mentor-reviewed workflow loops for local agents such as Hermes Agent, plus a shared experience dataset

What Agent Apprenticeship does

Agent Apprenticeship runs an apprentice agent through workflow loops on real-world tasks, with a mentor agent, the user or a human expert evaluating the work. Each run produces an Experience Compilation, a reusable record of the work that can feed future agents. It works with local agents including Codex, Cursor, Claude Code, OpenClaw, OpenCode, Hermes Agent and custom agents, alongside different model providers.

The command-line tool is apprentice, with commands for init, settings, doctor, configure, run and watch. Three modes are offered: Autonomous, Expert-Led and Organization Custom. Mentor model provider keys for OpenAI, Anthropic, Gemini or OpenRouter go in ~/.agent-apprenticeship/.env.local. A published seed dataset includes 500+ curated tasks, 495 reusable lessons, 1000+ execution traces and 505 experience compilations. Users can contribute learning signals back to the ecosystem.

Key features

Apprentice agent loops evaluated by a mentor agent, the user or a human expert
Experience Compilation output for every completed run
Works with Codex, Cursor, Claude Code, OpenClaw, OpenCode, Hermes Agent and custom agents
Autonomous, Expert-Led and Organization Custom modes
apprentice CLI with run, watch, doctor and configure commands
Seed dataset with 500+ tasks, 495 lessons and 1000+ execution traces

When to use it

Run a Hermes Agent task through a mentor review loop and keep a record of the result
Collect agent execution traces as training signals
Use shared ecosystem signals to improve your own agent

Who it is for: Developers and researchers who want agent runs to produce reviewed, reusable experience data.

How it fits with Hermes Agent

Hermes Agent is listed among the local agents Agent Apprenticeship supports, and the repository carries a hermes-agent topic.

How to install Agent Apprenticeship

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

npx agent-apprenticeship init
npm install -g agent-apprenticeship
apprentice init

Requirements: A supported local agent, and API keys for a mentor model provider (OpenAI, Anthropic, Gemini or OpenRouter) stored in ~/.agent-apprenticeship/.env.local

FAQ

What is Agent Apprenticeship?

Agent Apprenticeship is a framework where an apprentice agent does real-world tasks in workflow loops and a mentor agent or human evaluates the work. Completed runs become reusable experience records and training signals.

Does Agent Apprenticeship work with Hermes Agent?

Yes. The README lists Hermes Agent among the supported local agents, alongside Codex, Cursor, Claude Code, OpenClaw, OpenCode and custom agents.

How do I install Agent Apprenticeship?

Run npx agent-apprenticeship init, or install globally with npm install -g agent-apprenticeship and run apprentice init. Then use apprentice doctor to check your setup.