Built-in tools the OpenDunes gateway runs on your behalf, server-side, before a response reaches you.
Not yet available. Server tools are on the roadmap and not yet available.
Server tools extend the OpenDunes gateway with capabilities that run entirely server-side — your application sends a standard chat-completions request, and the gateway handles tool invocation, result injection, and re-prompting automatically. You do not need to implement a tool-call loop in your own code.
When you enable one or more server tools on a request, the gateway intercepts any tool-call decisions the model makes, executes the corresponding server-side function, injects the result back into the conversation, and continues generation — all within a single API call. Your application receives the final model response as if no tool call ever happened.
This is different from function calling, where your code owns the execution loop. With server tools, OpenDunes owns the loop.
Pass a tools array in your request body exactly as you would for standard function calling, but use the reserved namespace opendunes.* for built-in tools:
The gateway resolves the tool, runs it, and returns the completed response. The final message in choices[0].message is the model's answer after seeing the tool result.
| Tool | Description |
|---|---|
| Web Search | Live web search injected into the conversation |
| Web Fetch | Fetch and extract content from a URL |
| Datetime | Current date, time, and timezone resolution |
| Image Generation | Generate images and return them as message content |
| Apply Patch | Apply a unified diff to a code block |
| Advisor | Route sub-questions to a specialist model automatically |
| Subagent | Spawn an autonomous sub-task agent inline |
Each server tool invocation counts toward your balance. Tool execution cost is itemized separately from the model's token cost in your usage dashboard. You can set a per-request tool-spend cap to prevent runaway loops.