🧭 Claude Max and Team Subscriptions Now Include Monthly API Credits

As of October 7, 2026, Anthropic began bundling monthly API credits into its Claude Max and Team subscription tiers. The change means that paying subscribers can call the Claude API directly — without setting up a separate billing account or API key — up to a monthly credit allowance that resets each billing cycle. The credits work across all Claude Platform API models and do not roll over if unused.

Credit allocation by plan

Claude Max 5×    → $100 / month in API credits
Claude Max 20×   → $200 / month in API credits
Claude Team      → $500 / month pooled across the workspace

What the credits do — and what they don't

• Cover: Any API call to models on the Claude Platform (Sonnet 5.5, Haiku 5.5, Opus 5.5, etc.) — including prompt caching reads, batch API, the Tools API, and image/PDF input.
• Do not cover: Interactive claude.ai conversations, Claude Code chat-turn limits, or claude.ai artifacts. These remain subject to the existing plan message limits.
• No rollover: Unused credits expire at the end of each billing period. There is no accrual across months.
• Team pooling: On the Team plan the $500 is shared across all seats — useful for small engineering teams where one or two members drive most of the API usage.

🧭 Models API Gains capabilities.server_tools — Programmatically Check Web Search and Code Execution Support

On October 6, Anthropic extended the Models API with a new capabilities.server_tools object that lets developers programmatically inspect whether a given Claude model supports built-in server-side tools — currently web search and code execution. Rather than hard-coding which models support which tools and updating your code every time Anthropic ships a new model, you can now query the API at startup and branch on what you find. This follows two earlier additions in the same release cycle: capabilities.thinking.types.disabled (October 5) and a line field for model family grouping (October 1), pushing the Models API toward a fully machine-readable capability manifest.

The new fields

# GET https://api.anthropic.com/v1/models/{model_id}
{
  "id": "claude-sonnet-5-5-20260901",
  "display_name": "Claude Sonnet 5.5",
  "capabilities": {
    "server_tools": {
      "web_search":       true,
      "code_execution":   true
    },
    "thinking": {
      "types": {
        "disabled":  false,
        "enabled":   true,
        "streaming": true
      }
    },
    "line": "claude-5-5"
  }
}

Practical use: runtime capability guards

import anthropic

client = anthropic.Anthropic()

def model_supports_web_search(model_id: str) -> bool:
    """Check at startup — cache result, don't call per request."""
    info = client.models.retrieve(model_id)
    return (info.capabilities.server_tools or {}).get("web_search", False)

MODEL = "claude-sonnet-5-5-20260901"
USE_WEB = model_supports_web_search(MODEL)

tools = []
if USE_WEB:
    tools.append({"type": "web_search_20250305"})
    # add your web-search tool config here

response = client.messages.create(
    model=MODEL,
    max_tokens=1024,
    tools=tools,
    messages=[{"role": "user", "content": "What happened in AI this week?"}],
)