agent-apprenticeship
The living ecosystem where AI agents complete tasks through workflow loops, improve through iterative execution, are evaluated by mentor agents or humans in the loop, and turn completed work into reusable work experience and data to improve future agents.
What is it?
What it is
Agent Apprenticeship is an open‑source, command‑line ecosystem that lets AI agents perform real‑world tasks in iterative workflow loops, receive mentorship from other agents or humans, and turn each execution into reusable learning signals and experience compilations.
Why it exists
It was created to provide a compounding infrastructure where economically valuable agent work generates training data, improves future agents, and allows users to share and consume that collective experience—making agent work more productive, transparent, and economically meaningful.
Who should use it
AI developers, data scientists, and product teams looking to build, test, and iterate on autonomous agents in real-world workflows; researchers in reinforcement learning and agent-based systems; organizations wanting to capture and reuse agent work experience for continuous improvement.
Who should avoid it
Users without any programming or command-line experience; those who prefer fully managed cloud services without local control; individuals who cannot or do not want to handle API keys and environment configuration; teams lacking the resources to run and monitor iterative agent loops locally.
How it works
A quick walkthrough in plain English
How agent-apprenticeship works
Step 1 of 3
You interact with it
Open agent-apprenticeship, send a request, or connect it to your stack.
Features
Advantages
- Open-source MIT-licensed platform for community-driven development
- Comprehensive dataset enabling robust agent training
- Scalable infrastructure for long-horizon and specialized tasks
- Economic value tracking for agent work
- Reusable learning signals improve ecosystem-wide performance
- Supports diverse agent models and custom configurations
- Iterative workflow loops enhance agent learning efficiency
- Public ecosystem access for shared knowledge and resources
- Local deployment options for privacy and customization
Disadvantages
- Complex setup requiring API key configuration for model providers
- Dependency on external model APIs may incur costs or rate limits
- Technical expertise needed for custom agent configuration
- Learning curve for leveraging ecosystem experience compilation
- Potential variability in task quality based on seed dataset scope
- Limited real-time feedback mechanisms for agent performance
Installation
FAQ
How do I get started with Agent Apprenticeship?
You can initialize the environment by running `npx agent-apprenticeship init` or by installing it globally via `npm install -g agent-apprenticeship` and then running `apprentice init`.
Which AI agents are supported as Apprentice Agents?
The platform supports Codex, Cursor, Claude Code, OpenClaw, OpenCode, Hermes Agent, and Custom agents (where you can provide your own command template).
How do I configure API keys for Mentor Model Providers?
You can store your keys (such as OPENAI_API_KEY or ANTHROPIC_API_KEY) in the `~/.agent-apprenticeship/.env.local` file or export them as shell environment variables for the current session.
What are the different Apprenticeship Modes available?
The system offers three modes: Autonomous, Expert-Led, and Organization Custom.
How can I use previous agent experiences to improve new tasks?
You can use the `apprentice learn install <experience_compilation_path>` command to install Runtime Training from a prior experience, which can then be utilized in future agent runs.
Can I control the maximum number of iterations for an agent loop?
Yes, you can configure the maximum loop depth using `apprentice settings` or set a one-off limit for a terminal session using `export AA_MAX_ITERATIONS=X`.
Featured in Videos
YouTube tutorials and walkthroughs for agent-apprenticeship
Alternatives
Similar projects ranked by category, topics, and text overlap.
