Documentation
Extending bot capabilities with external integrations.
Tools let your bot take actions beyond generating text. A bot can call HTTP APIs, trigger webhooks, or connect to MCP servers. When the AI decides it needs to use a tool, it generates the required parameters and Neuron executes the call.
The most common tool type. Configure an endpoint URL, HTTP method, headers, and body template. Define a parameters schema so the AI knows what arguments to provide. The bot will call the API and use the response in its reply.
Example: Weather API tool
{
"name": "get_weather",
"description": "Get current weather for a city",
"type": "http_api",
"config": {
"url": "https://api.weather.com/current",
"method": "GET"
},
"parametersSchema": {
"type": "object",
"properties": {
"city": { "type": "string", "description": "City name" }
},
"required": ["city"]
}
}Webhook tools receive data from external services. When you create a webhook tool, Neuron generates a unique URL. External services POST data to this URL, and the bot can process and respond to it.
MCP (Model Context Protocol) tools connect to external MCP servers. Provide the server URL and Neuron will discover available tools automatically. This is useful for connecting to complex services that expose multiple capabilities.
Tools support multiple auth types: none (public APIs), api_key (key in header or query), bearer (Bearer token), basic (username/password), and oauth2 (OAuth 2.0 flow). Configure auth when creating or editing a tool.
Store sensitive values like API keys as secrets. Secrets are encrypted at rest and can be referenced in tool configurations. Add secrets from the tool's detail page.
Secrets are write-only — once saved, you cannot view the original value. You can only replace or delete them.
Before deploying, test your tools from the tool detail page. Provide sample arguments and Neuron will execute the tool and show you the response. This helps catch configuration errors early.