agentproto models

agentproto models [adapter] [--json]

Lists every model declared by installed agent adapters, annotating each with its provider and whether a provider API key is configured (✓ runnable, or ✗ missing key). Prices — when known — come from @agentproto/model-catalog ($ per 1M input/output tokens).

The point: --model anthropic/claude-opus-4-8 fails at runtime with "no ANTHROPIC_API_KEY" if you haven't set a key. This surfaces the issue up front so you pick a model your environment is ready for.

Flags

Output

Without --json, prints one adapter block per installed adapter with its status, then one model per line:

mastra-agent  (installed)
  ✓ anthropic/claude-opus-4-8               anthropic  $15/$75 per 1M
  ✓ openrouter/z-ai/glm-5.2                 openrouter $0/$0 per 1M
  ✗ openai/gpt-4.1                          openai     $10/$30 per 1M

✗ = no provider key. Set one:
    agentproto auth provider set openai <api-key>

With --json, emits:

{
  "adapters": [
    {
      "adapter": "mastra-agent",
      "status": "installed",
      "models": [
        {
          "id": "anthropic/claude-opus-4-8",
          "provider": "anthropic",
          "runnable": true,
          "inputPer1M": 15,
          "outputPer1M": 75
        }
      ]
    }
  ]
}

Provider key sources

A model is marked runnable (✓) when its provider has a key available from either:

1. The current process environment — ANTHROPIC_API_KEY, OPENAI_API_KEY, OPENROUTER_API_KEY, etc.
2. The stored provider key file (agentproto auth provider set …).

Examples

# All adapters, all models
agentproto models

# One adapter in detail
agentproto models mastra-agent

# Machine-readable
agentproto models --json | jq '.adapters[].models[] | select(.runnable) | .id'

See also

• auth.md — set provider keys with agentproto auth provider set
• run.md — pass --model to a one-shot turn
• chat.md — pass --model to an interactive session
• concepts/adapters.md — how adapters declare models