https://www.kryptonhq.com/docs/tutorials/Recent content in Tutorials on Krypton RuntimeHugoen-ushttps://www.kryptonhq.com/docs/tutorials/first-agent/Mon, 01 Jan 0001 00:00:00 +0000https://www.kryptonhq.com/docs/tutorials/first-agent/<p>This guide deploys a custom agent into a running Krypton cluster, verifies the gateway route, and shows where scaling signals come from. Krypton treats your container as a black box that speaks A2A, MCP, or plain HTTP.</p> <h2 id="ports--endpoints--the-two-minute-mental-model">Ports &amp; endpoints — the two-minute mental model</h2> <p>Krypton exposes <strong>two</strong> HTTP services. Talk to the right one:</p> <table> <thead> <tr> <th style="text-align: right">Port</th> <th>Service</th> <th>What lives here</th> </tr> </thead> <tbody> <tr> <td style="text-align: right"><strong>8080</strong></td> <td><strong>Gateway</strong></td> <td>All agent traffic. Anything under <code>/v1/agents/{ns}/{name}/*</code> is reverse-proxied to your pod — the protocol RPC at <code>/</code>, the A2A card at <code>/.well-known/agent-card.json</code>, OAuth callbacks at <code>/oauth/...</code>, MCP SSE streams, anything. The gateway strips the <code>/v1/agents/{ns}/{name}</code> prefix; your container sees the original sub-path.</td> </tr> <tr> <td style="text-align: right"><strong>8090</strong></td> <td><strong>Control plane</strong></td> <td>Web UI and introspection APIs (<code>/v1/agents</code>, <code>/v1/agents/{ns}/{name}/mcp/tools</code>, <code>/v1/agents/{ns}/{name}/status</code>). Operator tooling only — <strong>never</strong> the path your clients use to invoke an agent.</td> </tr> </tbody> </table> <p>Rule of thumb: if it&rsquo;s a normal A2A / MCP / HTTP client, point it at <code>:8080</code>. If it&rsquo;s a browser or <code>kubectl</code>-adjacent tool, <code>:8090</code>.</p>https://www.kryptonhq.com/docs/tutorials/first-llm/Mon, 01 Jan 0001 00:00:00 +0000https://www.kryptonhq.com/docs/tutorials/first-llm/<p>Krypton lets you serve an LLM the same way you manage the rest of your cluster: declare a resource, let the controller create the workload, and send traffic through the gateway.</p> <p>Krypton serves Hugging Face GGUF models with <a href="https://github.com/ggerganov/llama.cpp">llama.cpp</a>. A <code>Model</code> resource points at a repo and file, the controller creates a Deployment and Service, and Krypton exposes the model through OpenAI-compatible endpoints:</p> <ul> <li><code>GET /v1/models</code></li> <li><code>POST /v1/chat/completions</code></li> <li><code>POST /v1/completions</code></li> <li><code>POST /v1/embeddings</code></li> </ul> <p>Any OpenAI SDK can use the gateway as its <code>base_url</code>.</p>