Agent Registry
Agent Registry
Section titled “Agent Registry”The agent registry lets you manage all your AI agents in one place. Agents are automatically registered when the SDK sends the first run_start event, but you can also create them manually.
/agents — Agent list
Section titled “/agents — Agent list”View all registered agents with:
- Agent name and framework
- Total run count
- Last active timestamp
- Budget status
/agents/create — Register an agent
Section titled “/agents/create — Register an agent”Manually register a new agent with:
- Name — unique per user
- Description — what the agent does
- Framework — LangGraph, CrewAI, or custom
- Tags — JSON tags for organization
/agents/{agent} — Agent detail
Section titled “/agents/{agent} — Agent detail”View detailed agent information including:
- Run history and statistics
- Current configuration version
- Budget settings and usage
Config versioning
Section titled “Config versioning”Each agent can have multiple configuration versions. Add a new config at /agents/{agent}/configs with:
- Version identifier (unique per agent)
- Config — JSON configuration object
- Changelog — description of changes
- Active flag — mark as current version
This lets you track how your agent’s configuration evolves over time.
Budget management
Section titled “Budget management”Set a spending budget per agent at /agents/{agent}/budget:
- Define a budget in USD for the agent
- The ingest API checks cost against budget on every
run_endevent - If cost exceeds budget, the API response includes a budget warning
- View budget status on the agent detail page
Run replay
Section titled “Run replay”Replay any completed run at /runs/{run}/replay. This shows the full event timeline step by step, useful for debugging and understanding agent behavior.
Run comparison
Section titled “Run comparison”Compare two runs side-by-side at /agents/compare. Select two runs and see differences in:
- Event sequence
- Duration and timing
- Outputs and errors
Useful for A/B testing agent configurations or debugging regressions.