A tool is an action that an MCP client can invoke.
In 0mcp, each tool is generated from an API operation in your OpenAPI specification. When you import your API, 0mcp automatically creates a tool for every available operation, allowing AI clients to interact with your API through the Model Context Protocol.
For example:
| API operation | Generated tool |
|---|
GET /orders/{id} | get_order |
POST /tickets | create_ticket |
GET /users | list_users |
- Name
- Description
- HTTP method
- API path
- Parameters
- Request body schema
- Response schema
- Authentication requirements
AI clients use the tool’s name and description to decide which action to invoke, so clear and descriptive metadata leads to better tool selection.
From the Tools page, you can customize how each tool behaves without changing your API.
You can:
- Rename the tool
- Edit its description
- Enable or disable it
- Review its inputs and request body
- Configure authentication
- Set request timeout
- Limit maximum response size
- Allow or block HTTP redirects
These settings only affect how the tool is exposed through your MCP server. They never modify your upstream API.
Best practices
- Use clear, action-oriented names such as
get_order, list_customers, or create_ticket.
- Write descriptions that explain when the tool should be used, not just what endpoint it calls.
- Disable operations you don’t want AI clients to access.
- Carefully review destructive operations such as deleting resources or processing payments before exposing them.
Test tools that create, update, delete, send, or charge data with limited credentials before exposing them to production clients.
Publishing changes
Changes to tools are first saved as a draft.
Connected MCP clients continue using the current published version until you publish a new one. Before publishing, use the Playground to verify that your tools behave as expected. 
