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

# Anthropic

> Auto-instrument Anthropic Claude messages with zero-config tracing. Supports sync and async.

Valiqor automatically traces all Anthropic `messages.create` calls —
both sync and async. Every call is recorded as a span with model name,
token usage, cost, and full message content.

***

## Install

```bash theme={"system"}
pip install valiqor[anthropic]
```

This installs `valiqor` plus `anthropic>=0.18.0`.

***

## Zero-Config (Recommended)

Add a single import at the top of your app — all Anthropic calls are
automatically traced:

```python theme={"system"}
import valiqor.auto  # ← Add this line

import anthropic

client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain quantum computing"}]
)
print(message.content[0].text)
```

Every `messages.create` call is now traced with full metadata.

***

## Selective Instrumentation

If you only want Anthropic tracing (not other providers), use the
provider-specific function:

```python theme={"system"}
from valiqor.trace import anthropic_autolog

anthropic_autolog()

# Or using the namespace-style API:
from valiqor.trace import Anthropic
Anthropic.autolog()
```

Both are equivalent — they enable tracing only for Anthropic.

***

## Async Support

Async Anthropic calls are automatically traced:

```python theme={"system"}
import valiqor.auto
import anthropic

client = anthropic.AsyncAnthropic()

async def main():
    message = await client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=1024,
        messages=[{"role": "user", "content": "Explain quantum computing"}]
    )
    print(message.content[0].text)
```

***

## Content Block Handling

Anthropic responses use content blocks. Valiqor automatically extracts
text from both string content and structured content block lists:

```python theme={"system"}
import valiqor.auto
import anthropic

client = anthropic.Anthropic()

# Multi-block content is handled automatically
message = client.messages.create(
    model="claude-sonnet-4-20250514",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Summarize this document"},
                {"type": "text", "text": "Document content here..."}
            ]
        }
    ]
)
```

***

## What Gets Captured

Each traced Anthropic call records:

| Field           | Description                                           |
| --------------- | ----------------------------------------------------- |
| `model`         | Model name (e.g. `claude-sonnet-4-20250514`)          |
| `input_tokens`  | Input token count                                     |
| `output_tokens` | Output token count                                    |
| `total_tokens`  | Combined token count                                  |
| `cost`          | Estimated cost in USD                                 |
| `stop_reason`   | Why the model stopped (e.g. `end_turn`, `max_tokens`) |
| `messages`      | User and assistant messages                           |
| `duration_ms`   | Call latency                                          |
| `status`        | Success or error                                      |

***

## With Workflows

Combine with `trace_workflow` to group multiple Anthropic calls into a
single trace:

```python theme={"system"}
import valiqor.auto
from valiqor.trace import trace_workflow
import anthropic

client = anthropic.Anthropic()

with trace_workflow("claude-assistant"):
    # Step 1: Analyze
    analysis = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=1024,
        messages=[{"role": "user", "content": "Analyze this data: [...]"}]
    )

    # Step 2: Summarize
    summary = client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=512,
        messages=[
            {"role": "user", "content": f"Summarize: {analysis.content[0].text}"}
        ]
    )
```

Both calls appear as child spans under the `claude-assistant` trace.

***

## Disabling

To disable Anthropic tracing:

```python theme={"system"}
from valiqor.trace import disable_autolog

disable_autolog("anthropic")    # Disable Anthropic only
disable_autolog()               # Disable all providers
```

***

## Limitations

* **Streaming** is not currently instrumented — streamed responses are
  not captured in traces
* **Tool use** — Anthropic tool use responses are not explicitly parsed
  (tool\_use content blocks are captured as raw content)

***

## Next Steps

<CardGroup cols={2}>
  <Card title="Tracing Guide" icon="route" href="/workflows/tracing">
    Learn about traces, spans, workflows, and exporters.
  </Card>

  <Card title="Failure Analysis" icon="magnifying-glass" href="/workflows/failure-analysis">
    Run failure analysis on your traced Anthropic calls.
  </Card>
</CardGroup>