Prompt book

Connect twins by changing one line of code.

Start from staging, a preview server, or locally. Point API bases at the twin hosts. Seed the scenario. Read the trace.

1. Start a twin run

In the workspace: Twins → Twin Runs. Select services, pick short-lived or permanent, describe optional state, then Start twin run. Hosts look like https://slack.run-2048.twins.evalab.lifestyle.

2. Rewrite .env

SLACK_API_BASE_URL=https://slack.run-2048.twins.evalab.lifestyle
GMAIL_API_BASE_URL=https://gmail.run-2048.twins.evalab.lifestyle
CALENDAR_API_BASE_URL=https://gcal.run-2048.twins.evalab.lifestyle
GITHUB_API_BASE_URL=https://github.run-2048.twins.evalab.lifestyle

Four API base URLs now point to Evalab twins. MCP and CLI twins use the same run id.

3. Compare outcomes

Request activity lists method, path, status, and latency. Outcome, provider-call count, forbidden effects, and total latency sit on the same page. Define judging criteria against that evidence for every commit or PR.

Questions: hello@evalab.lifestyle