> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.ada.cx/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.ada.cx/_mcp/server.

# MCP tools: choose what the AI Agent reads from a tool result

Each MCP tool now has a **What the AI Agent reads** setting. **Full tool result** stays the default: the Agent reads the whole result your server returned. **Only the output variables** limits the Agent to the outputs you saved as variables, returned as one JSON object under their output names.

Use **Only the output variables** to keep a large or noisy result out of the conversation. The response size limit then applies to the outputs you defined, not to the whole result. The option needs at least one saved output.

The [Tools API](/reference/tools/overview) exposes the same choice as `mcp.llm_reads` on MCP tools, with values `full_result` and `mapped_outputs`.

To learn more, see [What the AI Agent reads](/docs/automation/tools/mcp-tools#what-the-ai-agent-reads).