Agents

List available coding agents for eval runs from the Lightsage CLI.

An agent is one runnable harness-and-model combination. Select an agent ID when starting an eval run instead of configuring the harness and model separately.

Command

Command	Description
lightsage agents list	List runnable agents.

List agents

lightsage agents list --output-format json

Example response:

{
  "data": [
    {
      "id": "claude-code:claude-opus-4-8",
      "name": "Claude Code · Claude Opus 4.8"
    },
    {
      "id": "codex:gpt-5.5",
      "name": "Codex · GPT-5.5"
    }
  ]
}

Use the id value as the --agent-id when creating eval runs:

lightsage eval-runs create \
  --eval-id eval_123 \
  --configuration-id config_123 \
  --agent-id claude-code:claude-opus-4-8