Custom Tools let you connect any REST API to Plinqx without writing code. Use the visual wizard to configure endpoints, authentication, parameters, and response mapping.Documentation Index
Fetch the complete documentation index at: https://hypha-docs.plinqx.app/llms.txt
Use this file to discover all available pages before exploring further.
Overview
Custom Tools are perfect for:- Internal APIs that aren’t available as providers
- Third-party services with REST APIs
- Legacy systems that need AI integration
- APIs without OpenAPI documentation
Creating a Custom Tool
The Custom Tool Builder is a 4-step wizard that guides you through the entire process.Step 1: Basic Information
Define the tool’s identity:- Name: Internal identifier (alphanumeric, underscores only)
- Used as:
custom__{name}in MCP tool lists
- Used as:
- Display Name: Human-readable name shown in the UI
- Description: What the tool does (shown to AI clients)
Step 2: Endpoint & Authentication
Configure the API endpoint:- HTTP Method: GET, POST, PUT, PATCH, or DELETE
- Base URL: The API’s base URL (e.g.,
https://api.example.com) - Path: Endpoint path (e.g.,
/v1/users) - Timeout: Request timeout in seconds (default: 30)
Authentication Types
Choose how to authenticate with the API:- None
- API Key
- Bearer Token
- Basic Auth
- Custom Header
No authentication required.
All credentials should be stored in the Vault and referenced using
{{vault.CREDENTIAL_NAME}} syntax. Never hardcode credentials.Step 3: Parameters
Define input parameters for the tool:Input Parameters
Add parameters that AI clients will provide:- Name: Parameter identifier
- Type: string, number, boolean, object, array
- Description: What the parameter is for
- Required: Whether the parameter is mandatory
Parameter Placement
Configure where parameters are sent:- Path Parameters: Variables in the URL path (e.g.,
/users/{userId}) - Query Parameters: URL query string (e.g.,
?limit=10&offset=0) - Headers: Custom HTTP headers
- Body: Request body (for POST/PUT/PATCH)
Step 4: Response Mapping
Map API responses to a consistent format:Response Mapping
Use JSONPath expressions to extract and transform response data:Store Visibility
Configure which AI platforms can discover this tool:- ChatGPT: Expose to OpenAI’s GPT Store
- Claude: Expose to Anthropic’s Claude
- Google Gemini: Expose to Google’s Gemini
Testing Your Tool
Before publishing, test your tool:Test thoroughly before publishing. Once published, tools are available to AI clients.
Publishing Tools
Once your tool is ready:Editing Tools
To modify an existing tool:- Open the tool from the Custom Tools list
- Make your changes in the wizard
- Click Save to update
Tool updates create new versions. Previous versions remain available for audit purposes.
Tool Status
Custom tools have three statuses:| Status | Description |
|---|---|
| Draft | Not yet published, not available to AI clients |
| Published | Available to AI clients via collections |
| Archived | Disabled, not available to AI clients |
Best Practices
Use Vault
Always store credentials in the Vault, never hardcode them
Clear Descriptions
Write detailed descriptions so AI clients understand tool purpose
Test First
Always test tools before publishing to production collections
Version Control
Use clear naming conventions for easy identification
Next Steps
Add to Collections
Add your custom tools to collections
Set Up Vault
Store credentials securely in the Vault