Your First Project

Get Mibo running against your AI system: create a project, connect your system, and verify the connection.

Note

Create an account at app.mibo-ai.com/register. You can sign up with email and password or with Google.

Your demo Agent

When you create your first project, Mibo automatically adds a Demo Support Bot Agent. It comes pre-populated with example test cases, past executions, results, logs, and traces, so you can explore the dashboard before connecting your own system.

The demo Agent is read-only:

• Browse its test cases, executions, traces, and logs to see what each screen looks like with real data.
• You cannot run tests, edit configuration, or send traces to it. Its endpoint is a placeholder, and Mibo disables the controls that would change or run it.
• It does not count toward your plan quotas (Agents, executions, or traces).
• Keep it as a reference. After your project has another Agent, you can delete the demo from the Agent menu.

When you’re ready to test your own system, create a project and the Agent you want to test using the steps below.

Create a project

A project is your top-level workspace in Mibo. All your Agents, tests, and results live inside it.

On an empty dashboard, Mibo opens the Start with your Agent form after it confirms that your project list is empty. If you already have a project, open the project selector and use its Create New Project control to open the same form.

In Start with your Agent:

1. Choose an Agent type: FlowiseAI, n8n, or Other agent.
2. Enter an Agent Name, such as “Customer Support Bot.”
3. Enter a Project Name, such as “Customer Support quality workspace.”
4. Optionally expand Optional project details to choose a Project Icon or add a Description (Optional). The Test Architect receives the project and Agent names, not their descriptions, so include important system details in your request.
5. Click Create project and agent.

Mibo creates the project and the Agent you selected, then opens the Agent configuration page. The First-test setup panel tells you what is missing. Fill in the required fields, then click Save & continue. Mibo opens Test Architect automatically.

If the project is created but Agent setup fails, the form keeps your entries and the button changes to Retry agent setup. Fix the reported error, then click Retry agent setup. This retries Agent setup without creating another project.

If Mibo cannot load your projects, it shows Couldn’t load your projects instead of the project form. Click Retry loading projects to try again. Mibo opens the first-project form only after confirming that your project list is empty.

Tip

Describe your system in your request. The Test Architect receives the project and Agent names, not their descriptions. Include important details about your system and the scenarios you want to test. For example: “A customer support agent that handles billing inquiries, account changes, and product returns for an e-commerce site.”

One project per AI system. If you have multiple agents or systems, create a separate project for each. This keeps your results clean and your metrics accurate.

Follow Getting Started

After you return to the dashboard, the Getting Started panel shows the next action for a non-demo Agent. When the dashboard is scoped to a specific non-demo Agent, the panel uses that Agent. In All agents or Demo, it asks Choose an agent to continue if your project has more than one non-demo Agent. With only one, it uses that Agent. Choose Validate now to create a test, run it, and inspect its result. Depending on your progress, the panel offers Continue setup, Create your first test, Run saved test, View run progress, Inspect your result, or Review failed run.

Choose Monitor production when you are ready to evaluate real interactions. Set up trace ingestion takes you to the Agent’s trace page. Create an API key, keep an eligible test Active, save it with Passive behavior set to Always run or Run when relevant, and then send traces. See Passive Testing for the full setup.

The dashboard also links to the seeded sample data. The Demo Support Bot is read-only, so use it to explore results rather than to create or run your own tests. If the panel cannot load Agents or tests, use Retry loading agents or Retry loading tests before troubleshooting your connection.

Email verification

You can create a project, connect an Agent, create tests, and create API keys before verifying your email. Google sign-in verifies the account automatically. Mibo requires a verified email before you run tests, ingest traces, or use AI features.

The dashboard keeps setup controls available and shows a Verify your email banner. If you use email and password:

1. Open the verification link Mibo sends after sign-up. It takes you back to Mibo when verification succeeds.
2. If you do not receive the message, click Resend verification email in the dashboard banner.
3. After opening the link, return to Mibo and click I’ve verified my email in the banner. Mibo refreshes your session so restricted actions can continue.

If a link has expired, return to the dashboard and resend it. Resending can be temporarily unavailable during the cooldown or a server rate limit, but I’ve verified my email remains available after you open the link. If you completed verification on another device, return to this dashboard and use the same confirmation action.

Add an Agent

An Agent is the connection between Mibo and your AI system. It tells Mibo where to send test inputs and how to read the responses. The first-project flow creates the Agent you select. To add another Agent, use the steps below. Each project can have multiple Agents, which is useful for testing different environments (staging vs. production) or different versions of the same system.

From your project, click Connect Agent in the Agents section. If the sidebar is collapsed, use the plus button in that section. It is labeled Connect Agent and opens the same form. You can also use Connect your agent in the Getting Started panel or demo banner. In the Connect Agent form, pick the option that matches your system: FlowiseAI, n8n, or Other agent. Choose Other agent for LangGraph, custom apps, or any agent with an API or traces. The documentation calls this connection the HTTP API path.

Enter an Agent Name, then click Continue to connection. Mibo opens the Agent configuration page. For active testing, enter the fields shown for your provider, including the endpoint URL, authentication, and Request Payload. HTTP API uses Response Text Path. n8n also shows Workflow ID, Webhook ID (Auto-detected), Response Node Name, and Response Text Path. Flowise extracts response text from the text field automatically. Click Test Webhook for n8n or Test Connection for other providers. Then click Save Changes. The optional Setup guide highlights each part of this page.

For n8n trace collection, see Collect traces during active tests.

n8nPush canonical traces from n8n workflows with the community node — Cloud + self-hosted.

FlowiseImportable Custom JS Function that posts canonical traces — Cloud + self-hosted.

HTTP APIAny system with a web endpoint.

Tip

Already instrumented with OpenTelemetry? Use OTLP — it works with n8n self-hosted ≥ 2.19, Flowise self-hosted with native OTel enabled, and any OTel-instrumented service. OTLP is a trace ingestion option, not an agent type. See Sending Traces for the OTLP vs. Your API decision matrix.

Tip

Describe this Agent in your request. The Test Architect receives the project and Agent names, not their descriptions. Include the work this Agent handles, such as: “Production Flowise agentflow that handles restaurant reservations, cancellations, and menu inquiries.”

What every Agent needs

For active testing, configure:

• A URL: the web address where your AI system receives inputs.
• Authentication (if required): a token, API key, or custom headers that your system expects.
• A message template: the JSON structure Mibo sends to your system, using {user_message} as a placeholder for the test input.

Note

If you only plan to use passive testing — where your system sends traces to Mibo instead of the other way around — you can create the Agent first and add active-connection details later. Passive testing needs an API key and a trace-ingestion setup; see API Keys & Trace Routing and Sending Traces.

Test your connection

On the Agent configuration page, fill in the required fields. Click Test Webhook for n8n or Test Connection for other providers. Mibo sends a test input using your current URL, method, authentication, and payload template. A CONNECTED badge confirms that the endpoint returned a successful response. Click Save & continue during first-test setup or Save Changes when editing the Agent later.

If the test fails, the form shows FAILED and Retry test. Fix the fields, then click Retry test. After the CONNECTED badge appears, click Save & continue during first-test setup or Save Changes when editing the Agent later.

Check these common issues:

• Wrong URL: double-check the address and make sure there are no typos.
• Missing authentication: if your system requires a key or token, make sure it’s entered correctly.
• System not running: verify that your AI system is deployed and reachable from the internet.
• n8n test webhook: activate the workflow manually, or use the production webhook URL for automated runs.

Create and run your first test

AI actions require a verified email. Before verifying, open the Agent’s Tests page and click Create new to write and save a test manually. Return after verification to run it.

1. In the Agent navigation, open Create tests with AI. Describe one behavior you want to check, such as how your support Agent handles a refund request.
2. Review the generated test. The test includes a scenario, an input, and checks called assertions that Mibo uses to evaluate the response. Click Save & run tests to save it and start a run, or click Save to prepare it and run it later from the test list. Save & run tests is available after you verify your email.
3. To run one test later, open Tests and click its play button. To run every active test, click Run All Tests.
4. Open the test case to inspect Last Result. Expand Execution Logs to see requests, responses, and errors. If your system returns or sends trace data, expand Last Run Trace or open Traces in the Agent navigation.

For status meanings and next actions, see Read the result status. If a run fails or stays pending, follow Recover from a run error before retrying.

Tip

Start with one test case. It verifies your connection and gives you a result to inspect before you run the full suite.

What’s next

Your project is set up and your Agent is connected. If you ran a test, learn how to read the quality dashboard. Then connect trace ingestion if you want deeper execution details or passive testing.

Creating test casesGenerate and refine test cases with the Test Architect.

Your quality dashboardRead quality metrics, assertion results, logs, and failure details.

Sending tracesChoose OTLP or Your API for trace ingestion and passive testing.