{
  "id": "page:docs-agent-mux-tutorials-mock-harness",
  "_kind": "Page",
  "_file": "wiki/docs/agent-mux/tutorials/mock-harness.md",
  "_cluster": "wiki",
  "attributes": {
    "nodeKind": "Page",
    "sourcePath": "docs/agent-mux/tutorials/mock-harness.md",
    "sourceKind": "repo-docs",
    "title": "Mock Harness",
    "displayName": "Mock Harness",
    "slug": "docs/agent-mux/tutorials/mock-harness",
    "articlePath": "wiki/docs/agent-mux/tutorials/mock-harness.md",
    "article": "\n# Mock Harness\n\nThe mock harness replaces the real CLI spawn with a scripted event stream. Use it for tests, CI, and demos where real credentials or network access are undesirable.\n\n## Enabling\n\nTwo equivalent switches:\n\n```bash\namux run claude --use-mock-harness --prompt \"test\"\n```\n\nor globally:\n\n```bash\nexport USE_MOCK_HARNESS=1\namux run claude --prompt \"test\"\n```\n\nProgrammatically:\n\n```ts\nawait client.run({\n  agent: 'claude',\n  prompt: 'test',\n  useMockHarness: true,\n});\n```\n\n## Scenarios\n\nThe mock emits a deterministic sequence of `AgentEvent`s chosen by scenario name:\n\n| Scenario          | What it emits                                               |\n| ----------------- | ----------------------------------------------------------- |\n| `text`            | A few `text_delta` events and a final `done`.               |\n| `tool-call`       | `text_delta` → `tool_call_start` → `tool_call_result` → `done`. |\n| `thinking`        | `thinking_delta` events before the final message.           |\n| `error`           | Emits an `error` event with `recoverable: false`.           |\n| `session-resume`  | Replays a prior session snapshot.                           |\n\nPick one with `--scenario`:\n\n```bash\namux run claude --use-mock-harness --scenario tool-call --prompt \"x\"\n```\n\nIf you invoke the package binary directly, the same agent scoping is available\nthere too:\n\n```bash\nmock-harness --agent claude --list\nmock-harness --agent claude --scenario tool-call\n```\n\nInteractive approval scenarios under `packages/agent-mux/harness-mock` now model real gating semantics rather than timer-only replay:\n\n- `interactive:yolo` auto-approves and then emits the post-approval output.\n- `interactive:prompt` waits for stdin before it emits the post-approval output and exits.\n- `interactive:deny` auto-denies, emits the denial path, and exits non-zero.\n- `interactive:timeout` waits for approval until the configured timeout, emits a timeout error, and exits non-zero.\n\n## Why use it\n\n- **CI**: No API keys, no flakiness.\n- **Adapter tests**: Validate `parseEvent` without invoking the real CLI.\n- **Integration tests**: Exercise hooks, session-save paths, and the event stream in isolation.\n\nSee [Harness & Mock](../reference/14-harness-mock.md) for the full event script format.\n",
    "documents": []
  },
  "outgoingEdges": [],
  "incomingEdges": [
    {
      "from": "page:docs-agent-mux-tutorials",
      "to": "page:docs-agent-mux-tutorials-mock-harness",
      "kind": "contains_page"
    }
  ]
}