Skills

Watch the walkthrough (jumps to the Install Skills section):

The BFFless skills give your AI assistant deep knowledge of the BFFless platform — deployments, pipelines, proxy rules, domains, chat, and more. Instead of reading docs yourself, just describe what you want and the assistant handles the rest.

Skills were originally built for Claude Code and are distributed as a Claude Code plugin, but they're plain markdown — also compatible with the open-source skills CLI, which lets you install them into any agent that reads from a .skills/ (or similar) directory.

Why Use the Skills?

The MCP Server gives your agent access to BFFless tools (create projects, manage deployments, etc.), but the tools alone don't teach how to use them effectively. The skills add:

• Domain knowledge — Concepts like aliases, pipeline handlers, proxy rule sets, and traffic splitting
• Guided workflows — Ask for "a contact form with email notifications" and the agent knows the exact handler chain to build
• Best practices — Patterns like "all proxy rules must go in a single rule set per project" and "assign rule sets to aliases for them to take effect"
• Reference for every feature — Chat setup, auth patterns, the use-bff-state React hook, GitHub Actions CI/CD, and more

Prerequisites

• An AI coding agent (Claude Code, or any agent that supports the skills CLI format)
• A BFFless instance with an API key (see MCP Server — Setup)
• The BFFless MCP server connected to your agent

If you haven't set up the MCP server yet, see the MCP Server page first.

Installation

Claude Code (plugin marketplace)

From within Claude Code, add the marketplace and install:

/plugin marketplace add bffless/skills
/plugin install bffless

Or via CLI:

claude plugin install bffless --scope user

Then reload plugins to activate:

/reload-plugins

Any agent (via npx skills)

The same repo works with the open-source skills CLI. From your project root:

# List available BFFless skills
npx skills add bffless/skills --list

# Install all skills
npx skills add bffless/skills

# Install a specific skill
npx skills add bffless/skills --skill chat

This clones the bffless/skills repo and copies the selected skills into your project so any compatible agent can read them.

Available Skills

Skill	Description
authentication	Cross-domain authentication using the admin login relay pattern, built-in /_bffless/auth endpoints, and cookie-based sessions
authorization	Two-level permission system with global and project roles
bffless	Knowledge about the BFFless platform, features, and setup
cache-and-storage	Cache rules for HTTP caching headers, storage backends, API keys for CI/CD, and user roles
chat	Adding AI chat to a site with full-page or popup widget layouts, skills, streaming, and message persistence
pipelines	Backend automation without code using handler chains
proxy-rules	Forward requests to backend APIs without CORS
repository	Browse deployments, manage aliases, view branches
share-links	Token-based sharing for private deployments
traffic-splitting	Distribute traffic across aliases with weights and rules
upload-artifact	GitHub Action for uploading build artifacts to BFFless
use-bff-state	React hook for managing server-side state with BFFless Data Tables and Pipelines

Usage (Claude Code)

Invoke the skill bundle with the /bffless slash command:

/bffless

This loads BFFless domain knowledge into your current conversation. From there, just describe what you want in plain language:

"Set up a chat endpoint with message persistence for my project"

"Create a proxy rule that forwards /api/weather to a weather API"

"Add a custom domain for my landing page project"

"Build a pipeline that validates a contact form, saves it to a DB Record, and sends an email"

The agent will use the MCP tools to execute each step, applying the correct patterns automatically.

When to load BFFless skills

Load them when you're about to do BFFless-related work:

Task	Example Prompt
Deploy & promote	"Show my latest deployments and promote the newest to production"
Build APIs	"Create GET and POST endpoints for a todos schema at /api/todos"
Set up chat	"Add AI chat to my project with Anthropic and message persistence"
Configure domains	"Map blog.example.com to my blog project's production alias"
Manage data	"Query the last 10 records from my contacts schema"
CI/CD setup	"Show me a GitHub Actions workflow that deploys to BFFless on push to main"
Debug pipelines	"Enable debug logging on /api/contacts and show recent logs"

You don't need to invoke /bffless for every message — once loaded, the knowledge persists for the rest of the conversation.

Example Session

Here's what a typical interaction looks like:

You:  /bffless
      I want to add a contact form to my project "acme/website" that saves
      submissions and emails me

Agent: I'll set that up. Let me:
       1. Create a "contacts" DB Record with name, email, and message fields
       2. Create a proxy rule set for the project
       3. Add a POST /api/contact pipeline with form_handler → data_create → email_handler
       4. Assign the rule set to your production alias

       [executes each step via MCP tools]

       Done! Your contact form endpoint is live at POST /api/contact.
       Here's a React component to use it: ...

Updating

Claude Code:

/plugin uninstall bffless
/plugin install bffless
/reload-plugins

npx skills: re-run npx skills add bffless/skills to pull the latest version.

Troubleshooting

Plugin not found after install (Claude Code)? Run /reload-plugins to activate newly installed plugins.

MCP tools not working? Skills provide knowledge but rely on the MCP server for tool execution. Verify your MCP connection:

1. Check that the bffless MCP server is configured in your agent
2. Ensure your API key is valid and has appropriate permissions
3. See MCP Server — Setup for connection details

Domain knowledge not loading? In Claude Code, make sure you invoke /bffless at the start of your conversation — the plugin loads context on demand, not automatically. With the npx skills CLI, confirm the skill files were copied into your project directory.

Related Features

• MCP Server — Raw MCP tool access for programmatic BFFless management
• Pipelines — Backend automation with chained handlers
• Chat — AI-powered chat with streaming and persistence
• Proxy Rules — API endpoints and request routing