multica-creating-agents

To create a Multica agent, use the multica agent CLI or POST /api/agents with required fields like name and runtime_id. The system enforces field validation and persistence rules before assembling the runtime payload used at claim time.

Multica agent creation requires fields such as name and runtime_id, alongside runtime_config and instructions. The validation process enforces these required fields and applies defaults to shape the runtime payload loaded by the daemon at claim time.

Secrets and custom_env variables are handled through secure channels including stdin, file, and inline inputs. The agent creation process gates these secrets with specific security and auditing guidance to ensure proper validation across create and update flows.

The mcp_config and custom_env fields receive distinct treatment during agent updates. The creation flow validates these inputs separately, ensuring mcp_config interactions and custom_env secrets are correctly persisted according to their specific field contracts.

Multica assembles the runtime payload at claim time by applying a separate skill-binding workflow to the persisted agent record. This payload includes the validated runtime_config, instructions, and mcp_config required by the daemon.

Yes, after creating an agent via the CLI or API, you can inspect the persisted agent record. Inspecting the record verifies that the field contracts, persistence rules, and runtime payload defaults were correctly applied.