Valet

One plugin: both skills, Valet's MCP server, and the publishing preference.

/plugin marketplace add valetdotdev/skills /plugin install valet@valet

Codex reads the same marketplace manifest:

codex plugin marketplace add valetdotdev/skills codex plugin add valet@valet

The two skills stay separate inside it, each with its own triggers, so only the one that fires pays its cost. The plugin adds ~510 tokens per session for both descriptions.

On Claude Code this installs the skills, declares the MCP server,andwires up the publishing preference.

npx skills add valetdotdev/skills --skill valet-publish -g npx skills add valetdotdev/skills --skill valet -g

Pick thisorthe plugin, not both — Codex loads skills from~/.agents/skills/and from installed plugins, so installing both puts two copies of the same skill in front of the model.

The hooks are a client extension and sit outside that specification, which defines no portable hook format — so a conformant client gets the skills and the MCP server, and the publishing preference only where its own hook system is wired up.

The plugin also declares Valet's own MCP server athttps://api.valet.dev/mcp, so installing it wires up the network path alongside the skills. The server registers the client itself. A normal publish starts the connector's OAuth flow when needed;anonymous: trueis reserved for an explicitly temporary public site.

The CLI and MCP connector keep separate credentials. Claiming an anonymous site signs the user into the browser and makes the site permanent, but it cannot authenticate the connector. The next account-first MCP call starts that client's OAuth flow.

It is there as the fallback the skills already describe, not as the main road.valet-publishdrives thevaletCLI, which publishes a whole directory from disk, updates a site later from anywhere, sets password access, and carries binary files — none of which the MCP tools can do. The server matters where the CLI cannot run: a sandbox whose outbound proxy refuses the install host, or a harness with no shell. Before the plugin declared it, reaching that fallback meant asking the user to add a connector by hand, mid-task, in a settings screen. Now it is already there.

Two files declare it, because no single format reaches both clients:

They must stay in agreement.make lint-valet-skillscompares them and fails on a server or URL that appears in one and not the other.

- SessionStartstates the preference once per session. This is the half that does the real work — deciding to render a page is usually the model's own call, and nobody says "publish this" out loud.
- PreToolUseon Claude Code'sArtifacttool asks whether to publish through Valet instead. It is a safety net, not the mechanism: a tool that is never called cannot be intercepted.

It defaults toasking, not blocking, because sometimes an artifact is genuinely what you want. SetVALET_PUBLISH_HOOKto change that:

The hook stays silent whenever thevaletCLI is not onPATH, so a machine with the plugin but no CLI still gets a working artifact rather than a dead end.

Launch Claude Code or another coding agent and use the/valetcommand:

$ claude > /valet build me an agent that reviews PRs for security vulnerabilities

The skill handles the full agent lifecycle — creating the workspace, writing the SOUL.md and skills, configuring channels, and deploying to the Valet Cloud.

Agents are defined by prompts, not code. A workspace is a directory:

my-agent/ SOUL.md # Agent identity skills/ # Agent capabilities <skill-name>/ SKILL.md channels/ # Event triggers (webhooks, cron, messages)

The Valet skill helps you create this structure, iterate on it, and deploy it — all from within your coding agent.

You can also install the Valet CLI directly:

curl -fsSL https://valet.dev/install.sh | sh valet auth login

irm https://valet.dev/install.ps1 | iex valet auth login

Existing Homebrew installations remain supported:

brew install valetdotdev/tap/valet valet auth login

To migrate from Homebrew later, uninstall the formula and run the direct installer for the platform. The CLI will then use direct self-updates.

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

Create crafted UI components inspired by the best 21st.dev design engineers.

Bring agent evaluations, observability, and synthetic test set generation directly into your IDE for free with Galileo's new MCP server

An MCP server to help AI assistants to answer questions and generate AccelByte Extend SDK code more effectively .

MCP server for AI Diagram Maker — generate beautiful software engineering diagrams directly inside Cursor, Claude Desktop, Claude Code, or any MCP-compatible AI agent

ALAPI MCP Tools,Call hundreds of API interfaces via MCP

AI-powered SVG animation generator that transforms static files into animated SVG components using the Allyson platform

MCP server that gives AI assistants on-demand access to 1,500+ amCharts docs, ~300 code examples, and 1000+ class API references.

APIMatic MCP Server is used to validate OpenAPI specifications using APIMatic. The server processes OpenAPI files and returns validation summaries by leveraging APIMatic’s API.

One shared context layer for AI agents and humans — live API specs, DB schemas, and versioned contracts across repos so every agent and teammate works from the same source of truth.