Claude Code Buddy: what happened to /buddy and what is safe now

The short answer

Anthropic shipped `/buddy` as a small terminal-only April Fools feature, then removed it. A maintainer confirmed the removal in the official Claude Code repository on April 9, 2026. Current Claude Code documentation does not list `/buddy` among its built-in commands.

If the command is missing, your installation may be working as designed. Do not downgrade Claude Code or paste an unofficial installer merely to restore the pet. First decide whether a community companion is worth the extra code, permissions, maintenance, and supply-chain risk.

Treat every replacement as third-party code. Confirm its origin, read the source, and inspect requested permissions. Test it away from important work, monitor changes, and remove it if the value is unclear.

What the original Claude Code Buddy was

The original feature added a small companion inside the terminal experience. Reports from the release described a generated creature, a card with traits or statistics, and reactions around the coding session. It was decorative and playful rather than a second coding agent.

The strongest first-party evidence is an issue in Anthropic's Claude Code repository. A user reported that `/buddy` worked in terminal version 2.1.90 but did not render in the desktop app. An Anthropic maintainer closed the request and described Buddy as a small April Fools feature removed from the latest release.

See the maintainer response in Anthropic's Claude Code issue 42677.

What the name does not identify

Search results now mix several different things. A result using the words Claude, Code, Buddy, companion, pet, or plugin does not inherit Anthropic's authorship, review, privacy terms, support, or security controls.

• The removed first-party `/buddy` feature inside a few Claude Code releases.
• Pixel-creature generators that recreate a visual companion or export an image.
• Community packages that alter configuration or install a custom command.
• MCP servers, hooks, status-line programs, skills, or plugins that add reactions and state.
• Development workflow tools that use Buddy as a product name but are not a terminal pet.
• Claude Desktop or Cowork components that use the word buddy for another internal or hardware function.

Check the current installation before changing it

Confirm which Claude Code surface and version you are using. Terminal, desktop, VS Code, and web do not expose every command in the same way. The live command menu is a better source than a screenshot from a short-lived release.

1. 1Run the official version command for your installed Claude Code client.
2. 2Inside an interactive terminal session, type `/` and search the available command list.
3. 3Compare the result with Anthropic's current command reference.
4. 4Check release notes and the official repository before treating a missing novelty command as a broken installation.
5. 5Use Anthropic's official install or update guidance for the supported client you intend to run.

Anthropic's current commands reference explains how to browse built-in commands, bundled skills, plugins, and MCP contributions.

Why an old restoration guide can be risky

A guide may have been accurate for a narrow release and still be a poor choice today. Downgrading can reintroduce fixed bugs, create incompatible configuration, or leave the client outside its supported update path. A copied command can also do much more than its surrounding page claims.

• A package runner can download and execute code immediately.
• An install script can change shell files, application settings, hooks, status-line commands, or authentication state.
• An MCP server may receive prompts, tool inputs, repository context, or environment details according to its configuration.
• A hook can run when Claude Code starts, stops, calls a tool, or processes another event.
• A replacement can store telemetry, companion state, logs, or identifiers outside the original client.
• A downgrade can conflict with newer desktop, plugin, model, permission, or settings behavior.

The domain hosting an instruction page is not proof that the command is safe. Shared content, advertisements, copied documentation, compromised repositories, and lookalike packages can all make untrusted instructions appear familiar.

Review a community replacement before running it

Treat a replacement like any other development dependency with access to an active workstation. The companion may look harmless while its installer or runtime receives broad access.

1. 1Identify the exact repository, package scope, publisher, maintainers, and release you are considering.
2. 2Inspect source for the installer, entry point, update path, hooks, MCP tools, status-line command, and uninstall process.
3. 3Open the package manifest and lockfile. Note install scripts, native binaries, downloaded assets, and network calls.
4. 4Review open issues, recent commits, release signatures, ownership changes, archived status, and unresolved security reports.
5. 5List every file, setting, environment variable, credential, permission, and network endpoint the tool can touch.
6. 6Decide what useful outcome justifies that access. A visual pet should not need unrelated repository, browser, wallet, cloud, or secret access.
7. 7Prefer a pinned, reviewed release over a floating command that downloads whatever is newest at execution time.
8. 8Keep a clean backup and a written uninstall plan before the first test.

Understand Claude Code extension surfaces

A companion can be implemented through several surfaces, and each has a different trust boundary. Use the official documentation for the surface a project uses rather than assuming all slash commands work alike.

Skill

prompt instructions and optional files or scripts that Claude can use when invoked.

Hook

a command triggered by a configured Claude Code event.

Status line

a program whose output appears in the terminal interface.

Plugin

a packaged extension that can contribute commands, agents, hooks, skills, or MCP servers.

MCP server

a separate process or service that exposes tools or resources to the client.

Direct configuration patch

code that edits Claude Code settings or other files without using a packaged extension path.

Anthropic documents how skills are structured and loaded in Claude Code.

Review Anthropic's permission-mode guidance before approving file, shell, or network actions.

Use a disposable test environment

Source review lowers risk but does not reveal every runtime behavior or future update. Test the exact pinned version in an environment that has no valuable credentials, repositories, personal files, active browser sessions, or production access.

• Use a disposable operating-system account, virtual machine, or development container where practical.
• Create a throwaway repository with no secrets or proprietary code.
• Keep cloud, package registry, Git, SSH, wallet, password manager, and browser credentials outside the test.
• Deny unnecessary network, file, shell, and environment access.
• Record files and settings before and after installation, then observe child processes and outbound connections.
• Restart the client and confirm whether the tool persists, updates itself, or changes behavior.
• Remove it and verify that hooks, servers, commands, files, and configuration entries are gone.

Back up configuration without exposing secrets

A configuration backup should help you recover, not create another credential leak. Copy only the files needed for comparison and store them in a protected local location. Inspect them for tokens, paths, account identifiers, or other sensitive values before sharing.

• Record the Claude Code version and installation method.
• List existing plugins, MCP servers, skills, hooks, status-line commands, and permission rules.
• Save a redacted copy of relevant settings before a tool changes them.
• Keep the original file permissions and document the restoration step.
• Do not upload authentication files or unredacted configuration to an issue, chat, or public repository.

Decide whether the companion earns its maintenance cost

A playful status indicator can make a terminal feel friendlier. It may also add distraction, startup work, updates, configuration drift, or another component to debug. The useful question is not whether Buddy is charming; it is whether the chosen implementation improves your experience enough to justify its access and upkeep.

• Does it communicate real status or only decorate the interface?
• Can it run without modifying authentication or broad settings?
• Can animation, sound, reactions, or notifications be disabled?
• Does it preserve terminal readability, accessibility, and reduced-motion preferences?
• Can the project be removed with no effect on normal Claude Code work?
• Is somebody actively maintaining it when Claude Code changes?

If /buddy stopped working

1. 1Confirm that normal Claude Code commands and coding tasks still work.
2. 2Check the current version and live command menu.
3. 3Compare your client with the current command reference and the official issue that documents removal.
4. 4Separate a missing removed command from a broader installation failure.
5. 5If the client itself is failing, follow current official troubleshooting guidance rather than a Buddy-specific downgrade.
6. 6Remove any partially installed replacement using its reviewed uninstall process, then inspect the configuration it changed.
7. 7Report reproducible client problems through Anthropic's supported channel without including secrets.

Start from the supported client in Better Design's Claude Code installation guide.

In summary

Claude Code Buddy was a playful, short-lived built-in feature, not a permanent coding agent. Anthropic removed `/buddy`, and current replacements are separate projects. Verify the entity, use official current documentation, review source and permissions, test in isolation, and keep an exit path.

Frequently asked questions