​
What Are Models, Rules, and Tools?

Continue agents are built from three main types of configuration:

Models

Language models that power different capabilities like chat, autocomplete, and agent mode

Rules

Guidelines and instructions that shape how the AI behaves and responds

Tools

MCP tools that provide additional capabilities like database access, web search, or custom functions
There are two places where you can define these configurations:

Local

Custom configurations you create and manage in your workspace or globally

Hub

Pre-built models, rules, and tools from the Continue community that you can import and use immediately

​
Local

Local configurations let you create custom models, rules, and tools that automatically apply to multiple agents, reducing duplication and ensuring consistency across your setup.

Global

Applied to all agents across all workspaces. Ideal for personal preferences, universal coding standards, or tools you use everywhere.

Workspace

Applied automatically to all agents when working in a specific project.Perfect for project-specific setups like TypeScript rules for web apps or the Playwright MCP tool.

​
Hub

Continue hub uses a slug in the format of owner/item-name to resolve blocks. For example, to use the Claude 4 Sonnet model, you’d reference it as anthropic/claude-4-sonnet. Import from the hub using the uses syntax alongside your custom configurations:
config.yaml
name: Team Agent
version: 1.0.0
schema: v1

models:
  - uses: anthropic/claude-4-sonnet
    with:
      ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }} # Use a hub secret

​
Organization

Organize your local configurations using these directories:

Models

.continue/models

Rules

.continue/rules

Tools

.continue/mcpServers

​
Working with Secrets

Models, and many MCP servers, require a secret for things like API keys. On the hub, you can configure secrets when adding a model or MCP server. This will use mustache notation to pass the secret, eg ${{ secrets.SECRET_NAME }} When configuring a local model or MCP server, you can use the same mustache notation for secrets which read from:

Global

.env file in ~/.continue/.env

Workspace

.env file at your project root

​
Overriding Properties

You can directly override properties using the override syntax:
config.yaml
name: myprofile/custom-agent
version: 1.0.0
schema: v1

models:
  - uses: myprofile/custom-model
    with:
      ANTHROPIC_API_KEY: ${{ secrets.MY_ANTHROPIC_API_KEY }}
      TEMP: 0.9
    override:
      roles:
        - chat

​
Advanced

​
Inputs

Models and MCP server authors can configure inputs that require the user to provide a secret value by defining a ${{ inputs.SECRET_NAME }} value. For example, here is how you could require that the user provide a value for the apiKey property on a model:
config.yaml
name: myprofile/custom-model
version: 1.0.0
schema: v1

models:
  - name: My Favorite Model
    # ... other model properties ...
    apiKey: ${{ inputs.SECRET_NAME }}
Users then map their secret to this input using a ${{ secrets.SECRET_NAME }} value that maps to a property name which matches the required input, e.g. SECRET_NAME.
config.yaml
name: myprofile/custom-agent
version: 1.0.0
schema: v1

models:
  - uses: myprofile/custom-model
    with:
      SECRET_NAME: ${{ secrets.SECRET_NAME }}

​
Next Steps

Now that you understand how models, rules, and tools work, explore: