Anthropic Server Tools

Server tools are tools where Anthropic's infrastructure executes the code. You enable the tool in your request; the server runs it and feeds results back to the model before the response reaches you. You never construct a tool_result block for these tools.

The current server tools are web_search, web_fetch, code_execution, and tool_search.

The server_tool_use Block

When a server tool runs, Claude's response contains a server_tool_use block with an id prefixed srvtoolu_:

{
  "type": "server_tool_use",
  "id": "srvtoolu_01A2B3C4D5E6F7G8H9",
  "name": "web_search",
  "input": { "query": "latest quantum computing breakthroughs" }
}

The result block follows immediately in the same assistant turn. Unlike client tool_use blocks, no further action from your application is required — execution is already complete by the time you see the response.

The pause_turn Continuation Pattern

Server tools run their own inner loop on Anthropic's infrastructure. If Claude is still iterating when the server-side iteration cap is hit, the response arrives with stop_reason: "pause_turn" rather than "end_turn".

Resume by re-sending the full conversation including the paused response:

if response.stop_reason == "pause_turn":
    messages = [
        {"role": "user", "content": original_content},
        {"role": "assistant", "content": response.content},
    ]
    continuation = client.messages.create(
        model="claude-opus-4-7",
        max_tokens=1024,
        messages=messages,
        tools=[{"type": "web_search_20260209", "name": "web_search"}],
    )

Include the same tools in the continuation to maintain functionality. You may modify the content before re-sending if you need to redirect or interrupt the agent.

ZDR Eligibility

The basic versions of web search (web_search_20250305) and web fetch (web_fetch_20250910) are eligible for Zero Data Retention (ZDR).

The _20260209 versioned tools with dynamic filtering are not ZDR-eligible by default because dynamic filtering relies on internal code execution. To use a _20260209 tool with ZDR, disable dynamic filtering:

{
  "type": "web_search_20260209",
  "name": "web_search",
  "allowed_callers": ["direct"]
}

"allowed_callers": ["direct"] restricts the tool to direct invocation only, bypassing the internal code execution step.

Domain Filtering

Server tools that access the web accept allowed_domains and blocked_domains to control which domains Claude can reach. Use one or the other — never both in the same request.

Domain filter rules:

No HTTP/HTTPS scheme in entries (example.com not https://example.com)
Subdomains are automatically included (example.com covers docs.example.com)
Specific subdomains restrict to only that subdomain
Subpaths are supported and match everything below the path
One wildcard per entry, after the domain part only (example.com/*, example.com/*/articles)
Invalid: *.example.com, ex*.com, example.com/*/news/*

Organization-level domain restrictions in the Console override request-level settings. Request-level domains can only further restrict, not expand, beyond the organization list. Mismatches return a validation error.

Unicode homograph attacks are a real risk: аmazon.com (Cyrillic ‹а›) looks identical to amazon.com. Use ASCII-only domain entries where possible.

Dynamic Filtering (_20260209 Versions)

The _20260209 tool versions use code execution internally to apply dynamic filters against results. Running a standalone code_execution tool alongside _20260209 web tools creates two separate execution environments, which confuses the model. Use one or the other, or pin both to the same version.

Streaming

Server-tool events stream as part of the normal SSE flow. The server_tool_use block and its result arrive as content_block_start / content_block_delta events, the same way text and client tool calls stream. All server tools support batch processing.