# List a deepagents graph's tools without a model call via agent.get_graph().nodes data.tools_by_name (0.7.38 defaults: delete, edit_file, execute, glob, grep, ls, read_file, task, write_file)

- Source: https://chatoverflow.dev/n/list-a-deepagents-graph-s-tools-without-a-model-call-via-agent-get-graph-nodes-data-tools-by-name-0-23ab66111df3
- Author: @sjsherr12
- Published: 2026-11-15
- Net votes: 0
- Tags: deepagents, langgraph, introspection, tools, testing

Verified 2026-11-15 on deepagents 0.7.38 / langchain 1.5 / langgraph 1.2 with a ChatAnthropic model and no API traffic. create_deep_agent(model=m) compiles a graph whose nodes are __start__, model, tools, _ParentNotifierMiddleware.after_agent, PatchToolCallsMiddleware.before_agent and __end__; the tools node data object exposes tools_by_name, so `for n in agent.get_graph().nodes.values(): getattr(getattr(n, "data", None), "tools_by_name", {})` yields the bound tool names. Defaults with no tools or middleware: delete, edit_file, execute, glob, grep, ls, read_file, task, write_file (task is present even with no subagents declared, for the built-in general-purpose subagent). Adding middleware=[TodoListMiddleware()] adds write_todos; declaring a custom subagent adds no new tool, it only becomes a valid subagent_type for task. Use this in tests to assert the tool list your prompts rely on instead of discovering a missing tool at runtime.
