Quick start: Skills, Docs, and Rules

See how skills, docs, and rules work together to make your agent more effective

This guide shows you how to set up Tessl and use docs, rules, and skills together to make your AI coding agent more effective. We'll build a simple Express.js REST API to demonstrate how these three components work together.

What you need:

• An AI coding agent (Claude Code, Cursor, Gemini, Codex, etc.)

• Node.js and npm installed

Setup Tessl

Step 1 of 5: Install the Tessl CLI

npm install -g @tessl/cli

See more detailed installation instructions

Step 2 of 5: Initialize Tessl

mkdir tessl-demo
cd tessl-demo
tessl init

This authenticates you with Tessl and configures your coding agent for MCP support (auto-detects Claude Code, Cursor, Gemini, Codex). See Custom agent setup for more options.

Setup your project

Step 3 of 5: Install documentation

tessl search express

Select the Express.js tile (e.g., tessl/npm-express 5.1.0) to install it.

What you get: Current Express.js documentation that your agent queries on-demand via MCP (Model Context Protocol).

Why it matters: Your agent's training data may be outdated (Express 4 vs Express 5). Documentation tiles provide accurate knowledge for your exact library version.

Step 4 of 5: Install rules

tessl install tessl-labs/quickstart-express-api-standards

What you get: Coding standards for error handling, validation, response formats, and security.

Why it matters: Your agent knows Express, but not your team's conventions. Rules teach it which validation library you use, your pagination defaults, and required security headers.

How it works: Rules are always loaded (eager push) - your agent follows them automatically in every interaction.

Step 5 of 5: Install skills

tessl install tessl-labs/quickstart-debug-api-endpoints

What you get: Systematic testing and debugging workflows (infrastructure → security → validation → functionality).

Why it matters: Skills capture proven procedures so your agent tests systematically instead of randomly - checking if the server is running before testing endpoints, verifying authentication before business logic, etc.

How it works: Skills load when relevant (lazy push). When you ask to test or debug an API, your agent follows the workflow instead of random patterns.

See it in action

Build your first endpoint

Ask your agent:

"Create a GET endpoint /api/hello that returns { message: 'Hello World' }"

What happens:

• ✅ Uses Express.js docs for current, correct code

• ✅ Follows your team's standards automatically (error handling, response format, security headers)

• ✅ Creates endpoint with tests following the systematic workflow (infrastructure → security → validation → functionality)

💡 Pro tip: For complex APIs, try Spec-Driven Development - define your API contract first, then let your agent implement it.

Test your API

Ask your agent:

"Test the /api/hello endpoint"

What happens:

The debugging skill activates, guiding your agent through systematic testing:

1. Infrastructure → Is the server running?

2. Security → Authentication, authorization checks

3. Validation → Input handling, error cases

4. Functionality → Happy paths, edge cases

Result: Comprehensive testing following proven procedures, not random patterns.

The power of combining all three

Type	When Used	Purpose
📚 Docs	On-demand (lazy pull)	Current library knowledge via MCP
📋 Rules	Always (eager push)	Your team's coding standards
🔧 Skills	When relevant (lazy push)	Systematic procedural workflows

Without Tessl tiles

❌ Outdated training data (Express 4 vs Express 5) ❌ Mixed version syntax ❌ Generic patterns, not your team's standards or workflows

With Tessl tiles

• ✅ Current documentation for your exact library versions

• ✅ Your team's conventions enforced automatically

• ✅ Your workflows instead of random patterns

Result: Current, consistent code following your team's standards.

Your agent does the right thing, automatically

Now as you continue developing, your agent will:

• Follow your team's standards - Error handling, validation, response formats - all applied automatically from the rules you installed

• Use proven workflows - Testing and debugging follow systematic procedures from skills, not random patterns

• Stay current - Query up-to-date documentation for the exact library versions you're using

• Remember your preferences - No need to repeat yourself in every request

Just ask your agent to build features, and it codes the way your team wants.

Creating your own tiles

Now that you've seen how tiles work together, you can create your own custom tiles with your team's standards and workflows. When you have team-specific coding standards, procedural workflows, or want to share knowledge across multiple repositories, creating custom tiles ensures consistent practices across your team.

For detailed guidance on creating tiles, see Creating tiles.

🎉 You're all set!

Explore more

Make your agents smarter with documentation

Creating skills

Autogenerating documentation

Distributing via registry