Give your coding agent the same evidence you see

Hand a live or recorded session to your coding agent for investigation, review its evidence in the player, and keep reusable tests and tasks in source control.

Give your coding agent a session ID so it can investigate the same screenshots, logs, telemetry, and app data you are reviewing in the player. You can follow the evidence visually while the agent works, then keep reusable checks and actions in source-controlled workspace definitions.

The player and agent use the same resident host. Agents access it through versioned ansight commands from their own terminal-capable environment.

Connect a workspace

Initialize or validate an Ansight workspace from the repository root:

ansight workspace init .
ansight test validate . --json

Workspace files define tests, tasks, triggers, Trends rules, sanitizers, and schemas in source control. The host discovers them from the connected repository; the player then exposes applicable Tests and Trends views.

To hand an existing workspace to another developer or agent, use workspace ZIP export and import. Export accepts either the repository path or a registered --app-id. Import restores the ansight/ files into a destination repository; review and validate them before linking that repository to the app. Host secrets, registrations, and session history are not part of the workspace ZIP.

Give an agent focused context

Start with the smallest stable identifiers:

ansight host status --json
ansight session list --connected --json
ansight session show <session-id> --json

Pass the selected session ID, app ID, or test ID into later commands. An agent can inspect screenshots, logs, telemetry, visual trees, artifacts, annotations, and guarded app tools without scraping the browser UI.

Keep evidence explicit

  • Session evidence remains local unless a person or command explicitly exports or shares it.
  • Repository tasks receive only their declared inputs and host defaults.
  • App tools remain subject to the SDK’s runtime guard and policy.
  • Host-local paths are never implied by selecting a file or session in the player.

Practical workflow

  1. Select the live or recorded session in the player.
  2. Copy its stable session ID or discover it with ansight session list.
  3. Ask the agent to investigate that ID with the relevant Ansight skill.
  4. Review the evidence or linked test trace in the player while the agent works.
  5. Keep reusable behavior in workspace tests, tasks, triggers, or Trends definitions.