Prompt Library and AI Tools: a better way to manage AI behavior

The Curiosity Team

• 2026-07-20

When AI behavior lives inside application code, with prompts buried in functions and tool logic scattered across files, iterating is slow. Changing a prompt means a deployment. Testing a tool means running the whole system. Understanding what the AI does means reading execution logic that was not written to be read.

Two new features in Curiosity address this: the Prompt Library and the AI Tools Library.

Prompt Library: manage AI prompts without touching code

The Prompt Library is one store of reusable prompt templates, separate from executable code. Each template supports variables with a \({VARIABLE} syntax (\), $, $), so the same template can serve many contexts.

The editor is split: the prompt on the left, a test panel on the right. The test panel extracts every variable from the template, shows an input for each, and renders the filled prompt as you type values. From there you can send it to an LLM and see the response, without leaving the admin.

The people closest to what the AI should say (product teams, domain experts, operations) can edit prompts, test them against live models and ship the change themselves. Prompt logic stays in one place, versioned, exportable and reusable across tools and agents.

AI Tools Library: give the AI functions to call

AI Tools are C# functions the workspace chat AI can invoke as tool calls: live data lookups, computations, calls to external services. The AI Tools Library is where those functions live, managed and tested in one place.

The editor has IntelliSense, autocompletion and compilation errors streamed from the server as you type, so a broken tool shows before a build. A Test Run panel beside the editor lists the exported functions, shows parameter inputs for the selected one, and runs it against the server, so you see exactly what the AI would get back.

Each tool has its own access control (admin only, all authenticated users, specific groups, or unrestricted). Usage metrics are built in: requests per second, P95 latency, error rate, calls and errors in the last hour.

Why separate these from code

The idea behind both is the same: AI behavior should be legible and editable without a full development cycle.

Prompts change often and need tuning. Tool logic changes less often, but needs to be testable and observable in production. Keeping both separate from application code means the people responsible for AI quality can do their own work, and the platform team is not the bottleneck every time something needs adjusting.

As an AI system grows, with more tools, more agents and more prompt variants, this separation stops being optional. The Prompt Library and the AI Tools Library are how Curiosity provides it.

If you are building on Curiosity, start with the developer docs, or get in touch.