Studio testing - Runiq.net Docs

Studio runs inside the same ASP.NET Core host as your agents. Use it to verify registration, try requests, and inspect model, tool, and retrieval behavior.

Enable Studio locally

After registering the agent and building the host, add this before app.Run():

Program.cs

using Runiq.AI.Core;

if (app.Environment.IsDevelopment())
{
    app.UseRuniqDashboard(options =>
    {
        options.Path = "/dashboard";
        options.Title = "Runiq Studio";
        options.Authentication(auth => auth.AllowAnonymous());
    });
}

Open /dashboard at your application's local address. This example enables anonymous access only in development. For other environments, follow Studio authentication.

1. Find the registered agent

Select Support Agent and confirm its identity, instructions, model, and attached capabilities. If it is missing, check the running application's AddRuniqServer registration.

For a CLI definition, select CodeReviewer or Claude Analyst from the corresponding guide. Metadata shows Codex CLI or Claude CLI with the configured model and requested reasoning effort. Both display attached Runiq tools. Installation and authentication failures appear when a request executes.

The prompt screenshot below shows a sample agent. Select the agent registered by your application; click the image to inspect it at a larger size.

2. Send a request

Start with a question within the agent's responsibility. Check whether it answers clearly or asks for missing details.

Use a request that needs a tool only after attaching that tool. A plain question does not need to produce a tool call.

3. Inspect the run

Check	What to look for
Instructions	The registered guidance matches the behavior you intended.
Model	The selected provider and model are correct.
Tools	The expected tool was requested with appropriate arguments.
Tool results	The operation succeeded and its output supports the answer.
RAG	When enabled, inspect retrieved passages and source references.
Completion	The response finished or exposes a useful failure to investigate.

Agent instructions shown in Studio

If an answer is unexpected, locate the first incorrect step. A retrieval problem needs a different fix from a tool failure or unclear instruction.

Repeat with a small evaluation set

Try a normal request, missing information, a request outside scope, and a failing tool or missing source when those capabilities are configured. Keep the same requests when changing instructions so you can compare behavior.

Studio supports manual inspection; opening a successful chat is not a complete automated evaluation.

Continue with the relevant guide

Organize growing agent definitions →

Inspect and configure RAG →

MCP tools and server setup →