Build with Claude Plugin Directory
Discover and install Claude Code plugins, agents, commands, hooks, and skills from one hub.
- Source repo
- davepoon/buildwithclaude
- Stars
- ★ 3.7k
- Last updated
- today
- License
- MIT
- Primary language
- Python
- FA score
- 46/100 · Major gaps
At a glance
- How it runs
- Works with
- Platform-specificClaude Code
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A Claude Code user wants to find and install a ready-made development agent or command by category.
- Not a fit if
- Teams seeking to install these plugins directly in a non-Claude Code environment
- Users who only need a local collection and not community discovery
- Source review
- 46/100 · Major gaps
What does this agent do, and when should you use it?
Build with Claude is a plugin marketplace and discovery platform for Claude Code. The repository maintains curated collections of agents, commands, hooks, skills, and bundled plugins, with marketplace commands for browsing and installation. Its web UI searches across plugin types and also indexes community plugins, MCP servers, and marketplaces. The collections are used within Claude Code workflows; the README also documents manually copying agent and command Markdown files into local Claude configuration directories. Direct use with other models or runtimes is not documented.
In Claude Code, users can run /plugin marketplace add davepoon/buildwithclaude to add the marketplace, search with /plugin search @buildwithclaude, and install with /plugin install <plugin-name>@buildwithclaude. The README lists 117 curated agents, 175 commands, 28 hooks, 26 skills, and 51 plugin packages; the web UI also indexes community plugins, MCP servers, and marketplaces. For manual installation, users clone the repository and copy Markdown files from plugins/agents-*/agents and plugins/commands-*/commands into ~/.claude/agents/ and ~/.claude/commands/. Hooks run on events such as tool calls or session start; agents can be invoked by context or explicitly, and commands use a / prefix.
- A Claude Code user wants to find and install a ready-made development agent or command by category.
- A team wants to add notification, Git, or formatting hooks to its workflows using
/plugin install. - A developer wants to search for plugins, skills, commands, and hooks in the web UI and copy install commands.
- An MCP user wants to browse a directory of servers for database, API, or tool connections.
- A Claude Code maintainer wants to manually copy agent or command Markdown files into local configuration directories.
How do you install or deploy this agent?
The recommended route is to add the marketplace and install a plugin in Claude Code:
/plugin marketplace add davepoon/buildwithclaude
/plugin install agents-python-expert@buildwithclaudeYou can also browse and search the directory at buildwithclaude.com. For manual installation of agents and commands, the README provides these steps:
git clone https://github.com/davepoon/buildwithclaude.git
cd buildwithclaude
find plugins/agents-*/agents -name "*.md" -exec cp {} ~/.claude/agents/ \;
find plugins/commands-*/commands -name "*.md" -exec cp {} ~/.claude/commands/ \;Restart Claude Code after copying files manually. The README does not specify any extra API keys or runtime requirements.
How do you use this agent?
After adding the marketplace in Claude Code, search for and install plugins with:
/plugin marketplace add davepoon/buildwithclaude
/plugin search @buildwithclaude
/plugin install <plugin-name>@buildwithclaudeAgents may be invoked by context or explicitly named in a prompt, for example @agent-security-auditor; commands use a / prefix, such as /commit, /docs, and /tdd. Hooks run on events such as tool calls or session start. The web directory supports browsing and filtering, viewing documentation and examples, and copying install commands.
What are this agent's strengths and limitations?
- Collects 117 agents, 175 commands, 28 hooks, 26 skills, and 51 plugin packages in one marketplace.
- Marketplace commands support search, individual installation, and examples for installing all agents, commands, or hooks by type.
- The web UI covers multiple plugin types and offers discovery for community plugins, MCP servers, and marketplaces.
- The README provides manual copy steps for agents and commands for workflows that do not use marketplace installation.
- Installation commands and plugin formats target Claude Code; direct use on other platforms is not evidenced.
- The manual installation steps cover agents and commands only, with no copy procedure for hooks, skills, or plugin packages.
- Collection and community-index counts are stated in the README, which does not describe update frequency or completeness.
- The README does not say whether web search requires an account or how often community index data is refreshed.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Build with Claude Plugin Directory This agent | 46 · Major gaps | Agent plugin / skillFree | ★ 3.7k | today | Python | Claude Code |
| Kimi Code CLI | 50 · Major gaps | CLI | ★ 7.9k | today | TypeScript | — |
| SkillSpec | 74 · Some gaps | CLIFree | ★ 708 | 2mo ago | Rust | Codex · Claude Code |
| OpenCode Goal Plugin | 73 · Some gaps | Agent plugin / skillFree + model costs | ★ 500 | 1d ago | TypeScript | — |
How does FollowAgents rate this agent?
Why each dimension lost points
The README gives marketplace add, search, and install commands and manual copy steps for agents and commands, but does not explain actual plugin permissions, data transfers, or removal and recovery; these areas receive 1. The MuAPI test shows that paid generation requires explicit confirmation, but this is evidence for one skill only, so user confirmation receives 1 for the repository-level product. The Atlas and GSC tests involve API keys and third-party services, while the supplied material lacks collection-wide guidance on sensitive data, dependency auditing, or external-effect controls. The MIT license and publisher identity do not provide adequate source-attribution detail.
README counts and product descriptions are not fully corroborated by package.json's collection description or the supplied test scope, so self-consistency receives 1. Workflows install Node dependencies and run validation, but one uses npm init and unpinned npm install without a lockfile in the shown material; dependency availability receives 1. Validation artifacts are configured, but there is no consistent user-facing failure guidance, so failure messages receive 1.
The README identifies development, operations, security, and other use cases and lists agents, commands, hooks, and skills, giving ordinary users a useful map of scenarios; audience and scenarios receives 2. Examples show invocation, but the supplied material does not systematically state each entry's permission boundaries, limitations, or trigger conditions, so those criteria receive 1. Marketplace and manual setup steps target Claude Code and receive 2; although the description also names Desktop, Agent SDK, and OpenClaw, the supplied installation instructions focus on Claude Code.
The README organizes categories, counts, quick start, installation options, examples, contribution guidance, and links, earning 2 for information architecture and examples. Manual installation covers only agents and commands; bulk copy may overwrite files and no uninstall or recovery instructions are shown, so install notes receive 2 and known limitations 1. A workflow checks agent filenames, but the supplied material does not establish stable naming across the collection or provide a change history; maintenance responsibility is only a general invitation to contribute. Those criteria each receive 1. The complete MIT text and its copyright and license terms are clear, earning 3.
The catalog and described one-click install-command copying can help users discover and add extensions, earning 2 for output usability. The repository offers multiple extension types and community discovery, showing practical discovery value; however, counts and coverage are mostly README claims and are not verified by the supplied files, so marginal value and cost-benefit receive 2.
README counts, platform coverage, and feature claims are not mapped item by item to a manifest or generated index in the supplied files. Test files support some individual skill behaviors but do not corroborate the product-wide claims. CI configuration shows validation steps, but two jobs mark failures advisory and no run results are provided. Separation of fact from inference is limited, so each verifiability criterion receives 1.
- This is a catalog of multiple types of agent extensions. Review each plugin's permissions and implementation before installation; tests for individual skills do not establish safety or quality for the full collection.
- The README provides bulk-copy installation steps but no uninstall or overwrite-recovery guidance. The supplied material also does not fully verify its category counts or coverage claims.