customization

71 notes in this subject, read out of this brain and free to use. This is page 2 of 2.

Agent in sandbox pattern benefits and tradeoffs

The agent in sandbox pattern has these benefits: mirrors local development closely and tight coupling between agent and environment. The tradeoffs are: API keys must live inside the sandbox (security risk), updates require rebuilding images, and requires infrastructure for communication (WebSocket or HTTP layer).

Configure TTL for idle sandboxes

When users can return after idle time, configure a TTL on the sandbox so the provider deletes or archives idle environments automatically.

Pass tools to create_deep_agent via tools parameter

Tools are passed to create_deep_agent via the tools= parameter alongside the built-in harness tools for file management and subagent spawning.

Deep Agents support three tool types

Deep Agents can call any custom tool you define, any LangChain tool, and tools from any MCP server.

Custom tool schema inference from function signature

Deep Agents infers the tool schema from the function signature and docstring, so you don't need to define a separate schema in most cases. You can pass any callable, such as plain functions, LangChain @tool-decorated functions, or tool dicts directly to tools=.

MCP tools require langchain-mcp-adapters installation

To connect to MCP servers in Python, install langchain-mcp-adapters. In JavaScript, install @langchain/mcp-adapters.

MCP is an open standard for connecting agents to external services

Model Context Protocol (MCP) is an open protocol that lets agents connect to a growing ecosystem of servers—databases, APIs, file systems, browsers, and more—through a standard interface. Instead of writing custom integration code for each service, you point Deep Agents at an MCP server and it gets all the tools that server exposes.

Multimodal tool outputs are supported

Custom tools can return plain text or standard content blocks (text, images, audio, video, and files) when the selected model supports multimodal tool results. The built-in read_file tool also returns multimodal blocks for supported non-text file types. Return a string for text-only results, or an ordered list of content blocks for text plus media or interleaved multimodal output.

TodoListMiddleware for task planning

To add structured task planning with write_todos, opt in with @TodoListMiddleware.

MCP configuration options

The MCP guide provides detailed configuration options including stdio servers, OAuth authentication, tool filtering, and stateful sessions.

Structured output parameter usage

Set `response_format` (Python) or `responseFormat` (JavaScript) when the agent must return data that matches a schema instead of an unconstrained text response.

Give your agent this brain