Schemas & Tools
AgenticAnts lets you save reusable structured-output schemas and tool definitions at the project level, then apply them in the Playground (and evaluations) — so you can test the exact structured-output and tool-calling behavior your application relies on.
Schemas and tools are created and managed inline from the Playground, and stored in a project-level library for reuse.
LLM schemas (structured output)
An LLM schema is a named JSON Schema that constrains a model's response to a defined shape (objects, arrays, required fields, enums). Use it when your app needs reliable, parseable output rather than free text.
- Give the schema a name and description, and define the JSON Schema.
- Apply it in the Playground's Structured Output section to force responses to match.
- Reuse the same schema across windows and evaluators.
LLM tools (function definitions)
An LLM tool is a named function definition — a name, description, and a JSON Schema for its parameters — that you can expose to a model so it produces tool calls.
- Define the tool's name, description, and parameters schema.
- Include it in a Playground window so the model can call it (the Playground records the tool call; it doesn't execute your function).
- Reuse tools across windows and evaluators.
Tools in the Playground let you test whether and how a model decides to call your functions and what arguments it produces — useful for validating tool-using agents before wiring up real execution.
Next steps
- Playground — where schemas and tools are applied.
- Prompt Management — save prompts that use them.