Upload your OpenAPI specifications to make your APIs usable by AI. Once uploaded and indexed, AI can find your operations, learn their interfaces, and execute them just like public APIs.
Toolcog supports:
Swagger 2.0 specs are not directly supported. Convert to OpenAPI 3.x first.
The upload starts immediately.
Toolcog validates your spec:
If validation fails, you’ll see error messages explaining what to fix.
Toolcog normalizes your spec:
$ref referencesEach operation is extracted into a tree-shaken spec:
AI generates semantic search phrases for each operation:
Intent phrases are embedded and indexed:
This process takes a few seconds for small specs, longer for large ones with hundreds of operations.
To update an existing API:
The new spec replaces the old one. Operations are re-extracted and re-indexed.
New operations are added to the catalog.
Operations not in the new spec are removed from the catalog.
Changed operations are updated with new schemas and intents.
Define authentication in components.securitySchemes:
components: securitySchemes: bearerAuth: type: http scheme: bearer apiKey: type: apiKey in: header name: X-API-KeyReference schemes in your operations:
paths: /users: get: security: - bearerAuth: [] # ...If your spec uses header parameters for auth instead of security schemes:
parameters: - name: Authorization in: header required: trueToolcog automatically detects and normalizes this to proper security schemes. See Overrides for manual control.
The summary field is used for intent generation:
paths: /customers: post: summary: Create a new customer description: Creates a customer record with contact information...Clear summaries → better intent phrases → better discovery.
Operation IDs appear in search results:
operationId: createCustomer # GoodoperationId: post_customers # Less clearTags help with filtering and organization:
tags: - name: customers description: Customer management operationspaths: /customers: post: tags: - customersAI uses descriptions to understand parameters:
parameters: - name: status in: query description: Filter by customer status (active, inactive, all) schema: type: string enum: [active, inactive, all]Examples help AI understand valid values:
schema: type: object properties: email: type: string example: "user@example.com"After uploading, add your API to catalogs:
Your API’s operations are now discoverable in that catalog.
Only you (and your organization) can see and use the API. Include in private catalogs for team use.
Anyone can discover and use the API. Include in public catalogs to share with the community.