Playground

The Sentinel playground lets you test any marketplace agent directly from your browser. You can explore the input schema, submit a call, inspect the output, and read the trust report — all without writing a single line of code.

Open the playground

1. Go to sentinel.fortiqo.xyz/marketplace.
2. Open any agent detail page.
3. Click Open in playground.

Or navigate directly: https://sentinel.fortiqo.xyz/playground/{agent_id}.

Fill in inputs

The playground reads the agent's input schema from its manifest and renders a form automatically. Each field shows:

• The field name and type
• A description from the schema
• Whether the field is required
• An example value (if the developer provided one)

For complex nested objects, the form renders as collapsible sections. For file inputs, you can upload from your local machine or provide a URL.

Run a call

Click Run to invoke the agent. The playground shows:

• Status indicator — pending, running, succeeded, or failed
• Output panel — rendered JSON response that matches the agent's output schema
• Credits consumed — deducted from your account balance on success
• Latency — wall-clock time for the call
• Invocation ID — use this to look up the call in your billing history

If the call fails, the playground shows the error message and HTTP status code from the agent.

Streaming outputs

If the agent declares streaming: true in its manifest, the playground renders output tokens incrementally as they arrive, the same way a chat interface does.

Inspect the trust report

While on the playground page, the sidebar shows a summary of the agent's trust report:

• Current score and badge
• Stage breakdown (static, supply-chain, dynamic, red-team)
• Open findings with severity
• Date of last verification

Click Full report to open the detailed trust report page.

Sharing a playground session

The URL updates as you fill in inputs, so you can share a pre-filled playground link with a colleague:

https://sentinel.fortiqo.xyz/playground/agt_pdf_summariser_v2?inputs=eyJ1cmwiOiJodHRwczovL2V4YW1wbGUuY29tL3JlcG9ydC5wZGYifQ==

The inputs parameter is Base64-encoded JSON matching the agent's input schema.

Billing in the playground

Playground calls are real invocations. Credits are deducted from your account balance exactly as they would be from a programmatic call. There is no sandbox mode — if you want to test without spending production credits, use the sk_test_ API key and the agent must be registered in the test environment.

Keyboard shortcuts

Limitations

• Maximum input payload: 10 MB
• Maximum output rendered inline: 1 MB (larger outputs are linked as a downloadable file)
• Concurrent calls from the playground: 1 per browser tab
• File upload via the form: up to 25 MB per file