> ## Documentation Index
> Fetch the complete documentation index at: https://docs.anpord.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Mock CLIs

> Test agent command use without external credentials or data

Define typed commands with Standard Schema, then attach the CLI to an eval:

```ts theme={null}
import { defineEval } from "anpord";
import { cli, command } from "anpord/cli";
import { z } from "zod";

const user = { id: "user_1", name: "Ada" };

const client = cli({
  name: "example",
  version: "1.0.0",
  description: "Manage users",
  commands: [
    command({
      name: "users get",
      description: "Get a user by ID",
      inputSchema: z.object({ id: z.string() }),
      outputSchema: z.object({ id: z.string(), name: z.string() }),
      options: { id: { description: "User ID", type: "string" } },
      handler: ({ id }) => ({ ...user, id }),
    }),
  ],
});

export default defineEval({
  name: "uses-users-cli",
  cli: [client],
  // cases, prompt, source, tasks, and trials
});
```

The same definition works with Codex, Claude, OpenCode, Pi, Gemini, Qwen,
Cursor, and FX. Anpord bundles each CLI into the task profile and puts its name
on `PATH`. Handlers use plain TypeScript and may return a value or promise;
inputs and outputs are validated against their schemas.

MCP tools and CLI commands can import the same fixture values and schemas. No
external organization IDs, service credentials, or test data are required.

Inspect exact calls in a validator:

```ts theme={null}
import type { Validator } from "anpord";

export const calledUsers: Validator = async ({ cli }) => ({
  message: "the agent ran users get",
  passed: (await cli.calls("example")).some(
    (call) => call.command === "users get"
  ),
});
```