Education
An environment provides activities, resources, constraints, and evaluation criteria. A student's agent helps them complete the activity, while an instructor's agent can evaluate submissions using a shared contract.
Experimental open pattern for agent-portable experiences
Bring your agent.
Enter the experience.
Do the work.
BYOAg allows people to bring their own AI agent into an external environment where the experience provides the context, artifacts, tools, rules, and interaction contract.
User-owned agent
Carries continuity, preferences, and working style.
BYOAg entry point
Publishes context, rules, artifacts, tools, and a contract.
Submission Feedback Evidence Continuation
Traditional AI integrations bring data, tools, and context into an agent-controlled environment.
BYOAg allows an agent to enter an environment-controlled experience. The agent remains the user's agent. The environment defines the work.
The core model
Four roles, one interaction contract
BYOAg separates who brings intelligence from who defines the experience, artifacts, and rules.
01
Human
Chooses the agent and authorizes the work.
02
Personal agent
Interprets the package and assists the user.
03
BYOAg environment
Publishes the work surface and evaluates results.
04
Artifacts, tools, rules, and contracts
Constrain, explain, and record the interaction.
How it works
A small lifecycle for agent entry
- 1 The environment publishes an entry point.
- 2 The user gives the entry point to their agent.
- 3 The agent retrieves a Prompt Package or Agentic Contract.
- 4 The agent performs the work and submits a result, receives feedback, or continues to the next interaction.
Why BYOAg
A pattern for shared work without forced agent ownership
The goal is not to guarantee perfect interoperability. It is to make the work surface explicit enough for different capable agents to participate.
Users retain continuity with their preferred agent.
Experiences do not need to own all intelligence.
Environments can define consistent rules and evaluation criteria.
Interactions can generate evidence of what was requested, attempted, and submitted.
The same environment can work with different capable agents.
Initial use cases
Places where the pattern can be explored first
Games and interactive experiences
An agent enters a game, trivia session, simulation, or branching story. Each submission can produce feedback, scoring, or the next scene.
Guided workflows
An environment provides structured forms, procedures, artifacts, and checkpoints while the user's agent helps them navigate the process.
Experimental status
BYOAg is currently an experimental pattern.
It is being developed through prototypes, working examples, architectural discussion, and an evolving draft specification. Developers, researchers, product designers, educators, and agent builders are invited to explore, question, and contribute.