EMAIL RUN: Execute evaluators against an agent that communicates over email. Cekura sends an email to the agent's address and simulates the full conversation, receiving replies via inbound webhook.

The agent must have an email_id configured, or you can supply agent_email as a one-time override. The agent_email field takes precedence if both are set.

Run-count: total conversations = len(scenarios) × frequency.

Inbound agents (agent.inbound=true): Cekura sends the first email to the agent. Outbound agents (agent.inbound=false): the response includes inbound_email — the address your agent should email first to start the conversation.

Polling for completion: the response is the result itself — read its top-level id (status: 'running'). Poll GET /test_framework/v1/results/{result_id}/ until status is completed, failed, timeout or cancelled. A result stays non-terminal until every run has ended, then becomes completed if any run completed — even if runs failed their checks or others errored. To gate, require success_runs_count == total_runs_count (failed_runs_count counts only completed runs that failed evaluation).