What Are Catalogs?

A catalog defines what API operations AI can discover and use. When AI searches for operations, it searches within a catalog’s scope. The catalog determines what’s findable.

The Standard Catalog

By default, Toolcog connects to the standard catalog:

https://mcp.toolcog.com

This catalog contains 100,000+ curated operations across hundreds of services—GitHub, Stripe, Google, AWS, and more. It’s everything you need to start using APIs immediately.

Why Create Catalogs?

Catalogs let you control what AI can discover:

Team-Specific Tools

Your engineering team might use:

• Internal APIs
• Specific GitHub repositories
• Select AWS services

A custom catalog includes only these, making AI more focused and efficient.

Private APIs

Upload your own OpenAPI specs to make internal APIs discoverable:

• Company services
• Backend APIs
• Internal tools

AI can discover and use them just like public APIs.

Security Boundaries

Limit what AI can access:

• Production vs. staging environments
• Customer-facing vs. internal operations
• Read-only vs. read-write operations

Custom catalogs enforce these boundaries.

Curated Subsets

Create focused catalogs for specific use cases:

• “Payments Catalog” with Stripe, PayPal, Square
• “Marketing Catalog” with Mailchimp, HubSpot, Salesforce
• “DevOps Catalog” with GitHub, AWS, Cloudflare

AI discovers only relevant operations.

The Composition Model

Catalogs are composed from sources. A source can be:

• An API — Include all or some operations from a public API
• Another catalog — Include operations from an existing catalog
• Your uploaded specs — Include your own APIs

Sources can be filtered to include only specific operations:

Source: GitHub API
  Filter: tag = "issues"
  Result: Only GitHub issue operations

Source: Stripe API
  Filter: endpoint = "POST /v1/*"
  Result: Only Stripe write operations

Multiple sources combine to form the complete catalog.

Public vs. Private Catalogs

Public Catalogs

Anyone can:

• Discover the catalog
• Connect to it via MCP
• Use the operations within

Useful for open-source tools, public APIs, community resources.

Private Catalogs

Only authorized users can:

• See that the catalog exists
• Connect to it
• Use operations within

Useful for internal tools, proprietary APIs, team resources.

Connecting to Catalogs

Your MCP URL determines which catalog AI uses:

https://mcp.toolcog.com                        # Standard catalog
https://mcp.toolcog.com/mycompany/internal     # Custom catalog
https://mcp.toolcog.com/mycompany/payments     # Another custom catalog

Each session is scoped to exactly one catalog. AI discovers only operations within that catalog.

Catalog URLs

Catalog URLs follow a consistent pattern:

https://mcp.toolcog.com/{owner}/{catalog}

• owner — The user or organization that owns the catalog
• catalog — The catalog’s handle (name)

For the standard catalog, no path is needed.

Use Cases

Internal API Portal

Upload your company’s OpenAPI specs. Create a catalog that includes them. Share the MCP URL with your team. AI can now discover and use your internal APIs.

Multi-Environment Setup

Create separate catalogs for:

• mycompany/production — Production APIs only
• mycompany/staging — Staging APIs for testing
• mycompany/development — Development APIs for building

Each environment has its own discovery scope.

Client-Specific Catalogs

Create catalogs per client:

• agency/client-a — Client A’s APIs and services
• agency/client-b — Client B’s different stack

AI working on Client A never discovers Client B’s operations.

Next Steps

• Creating a Catalog — Step-by-step catalog creation
• Adding Sources — Add and filter sources
• Uploading APIs — Add your own OpenAPI specs