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

# Context Editing

> Clear older tool results from Claude's server-side context with Anthropic context editing.

```python context_management.py theme={null}
from agno.agent import Agent
from agno.db.sqlite import SqliteDb
from agno.models.anthropic import Claude
from agno.tools.websearch import WebSearchTools

agent = Agent(
    model=Claude(
        id="claude-sonnet-4-5-20250929",
        betas=["context-management-2025-06-27"],
        context_management={
            "edits": [
                {
                    "type": "clear_tool_uses_20250919",
                    "trigger": {"type": "tool_uses", "value": 2},
                    "keep": {"type": "tool_uses", "value": 1},
                }
            ]
        },
    ),
    db=SqliteDb(db_file="tmp/context_management.db"),
    instructions="Use web search to answer with current, sourced information.",
    tools=[WebSearchTools()],
    session_id="context-editing",
    add_history_to_context=True,
    markdown=True,
)

if __name__ == "__main__":
    agent.print_response(
        "Search for AI regulation in the US. Make at least three searches, then summarize the latest developments."
    )

    run = agent.get_last_run_output()
    if run and run.metrics:
        print(f"Input tokens: {run.metrics.input_tokens:,}")

    if run and run.messages:
        for message in reversed(run.messages):
            context_data = (message.provider_data or {}).get("context_management")
            if context_data:
                edits = context_data.get("applied_edits", [])
                if edits:
                    latest_edit = edits[-1]
                    print(
                        f"Cleared tool uses: {latest_edit.get('cleared_tool_uses', 0)}"
                    )
                    print(
                        f"Cleared input tokens: {latest_edit.get('cleared_input_tokens', 0):,}"
                    )
                    break
```

The `clear_tool_uses_20250919` edit activates once the prompt contains more than two tool use/result pairs. It keeps the newest pair and clears older tool results. By default, Anthropic keeps the original tool-call inputs. Set `clear_tool_inputs=True` in the edit only when those inputs can also be removed.

Context editing runs on Anthropic's server before the request reaches Claude. The SQLite database stores Agno's session and run history independently of those server-side edits. Clearing content can invalidate the affected prompt-cache prefix.

Thinking blocks require a separate `clear_thinking_20251015` edit. When both strategies are used, list the thinking edit first. Context-edit metrics are finalized at the end of a response, including a streamed response.

See Anthropic's [context-editing documentation](https://platform.claude.com/docs/en/build-with-claude/context-editing) for strategy defaults and additional fields.

## Run the Example

<Steps>
  <Snippet file="create-venv-step.mdx" />

  <Step title="Install dependencies">
    ```bash theme={null}
    uv pip install -U "agno[anthropic,ddg,sqlite]"
    ```
  </Step>

  <Step title="Set your API key">
    <CodeGroup>
      ```bash Mac/Linux theme={null}
      export ANTHROPIC_API_KEY="your_anthropic_api_key"
      ```

      ```powershell Windows theme={null}
      $Env:ANTHROPIC_API_KEY="your_anthropic_api_key"
      ```
    </CodeGroup>
  </Step>

  <Step title="Save and run">
    Save the example as `context_management.py`, then run:

    ```bash theme={null}
    python context_management.py
    ```
  </Step>
</Steps>