There's a built-in agent inside Claude Code you can invoke just by telling Claude to use the claude-code-guide agent Save this... you'll need it It's not a local file in your project. It's a built-in subagent defined in Claude Code's system itself. It answers questions about Claude Code CLI,... the Agent SDK, and the Claude API. It uses Glob, Grep, Read, WebFetch, and WebSearch, but can't edit or write files. Its only job is to answer questions.show more

If you want richer output when learning a specific feature, change output-style to Learning or Explanatory inside /config

I didn't know about this one - super useful tip, thanks for sharing!

Did not know Claude Code had internal subagents like this---Definitely saving this!😍

This built in guide pattern is underrated. It gives consistent answers without forcing people to copy prompt templates into every repo.

Very cool! Will be giving this feature a spin

How is the UI on the right displayed?

Wow! It works, just tested it

Nice find. Built-in guide without write access? Safe by design.

been running on Claude Code for weeks and I'm still discovering hidden capabilities. it's like living in a house and finding a new room every time you lean on the wrong wall.

So Claude Code ships with its own helpdesk agent

While it's convenient to access the claude-code-guide agent, it's not entirely clear how it differs from local documentation in terms of reliability and customization. Depending on the project, a local file might still be preferable for certain use cases. Does the built-in agent offer any significant advantages over a well-maintained local file.

bro unlocked the tutorial npc inside claude

Yo, will you try out the guide on your next CC project?!

No way! How did you find this?

@nickcraske Because I'm a damn addict to Claude Code... hahaha I also read every single article and post from these guys @bcherny, @trq212 , @anthony

@bcherny @trq212 @Anthony So good. You’re adding fuel to my engines. Let’s go! 🙌

You don’t need to know about it. You can just ask Claude about Claude and he will use it

read-only by design. the guide agent has no write access because it doesn't need any. most people building internal agents give broad permissions then try to constrain later. scope it narrow from the start and expand only when required. anthropic baked this into the core product.

This is exactly the kind of built-in utility that makes learning Claude Code so much more efficient. A guide agent right there in the system? Smart. We're definitely bookmarking this for our next project. Thanks for the tip, @dani_avila7!

Hidden system agents like this change how we think about toolchains and autonomy.

Thanks for sharing with the community! I did notice CC calls it automatically at times, but explicitly instructing it using the tool is much more reliable getting most up-to-date information.

this is good to know. can't tell you how many times i’ve missed something simple because of not exploring the built-ins. appreciate the heads up!

interesting

the read only constraint is the smartest design decision here. most people building agent systems default to giving everything full access then wonder why it goes off the rails. having a dedicated guide agent that can search and read but never write means you get reliable answers without risking your codebase. I use a similar pattern where my AI agent spawns read only subagents for research tasks... containment by default is the way

Built in guide agents are underrated. Nice find

wait, so claude been hiding power moves in the system prompts