A browser conversation is a direct way to try a local language model before integrating it with an application. The llama.cpp server web UI accepts your prompt and displays the loaded GGUF model's answer.

llama-server serves the web UI from the same process that handles chat requests, so no separate Open WebUI installation is needed. The model chip identifies the active model; browser tools are optional and can be disabled for a plain text conversation with a small chat model.

A loopback listener keeps the UI accessible from the same machine without exposing prompts, responses, and inference capacity to other computers. Local access still has no API-key protection unless authentication is configured, and a remote listener needs authentication, network restrictions, and encrypted transport before exposure.

Steps to use the llama.cpp server web UI:

  1. Start llama-server with the local chat model on loopback port 8080.
    $ llama-server -m models/chat-model.gguf --alias local-chat --ctx-size 2048 --host 127.0.0.1 --port 8080

    This launch assumes llama-server is installed and a chat-tuned GGUF is already saved as models/chat-model.gguf. The web UI is enabled by default unless the server starts with --no-ui. The server terminal stays occupied while it runs.
    Related: How to start the llama.cpp server
    Related: How to set llama.cpp server host and port

  2. Check the single-model server's health from a second terminal.
    $ curl --fail --silent --show-error http://127.0.0.1:8080/health
    {"status":"ok"}

    The loaded server returns status set to ok. During loading, HTTP 503 makes curl exit nonzero.
    Related: How to check llama.cpp server health

  3. Open http://127.0.0.1:8080 in a browser on the server's machine.

    A new conversation shows Hello there and the Type a message… box. The local-chat chip matches the alias in the startup command.

  4. Click the plus button at the lower left of the message box.

    Its accessible name is Add files, prompts, tools or MCP Servers.

  5. Choose Tools from the menu.
  6. Clear the checkbox beside Browser for a conversation without browser tools.

    The group changes from 2/2 to 0/2. Browser tools can lead a small model to request a tool instead of answering a simple question.

  7. Enter the question What is two plus two? in the Type a message… box.
  8. Click the upward-arrow Send button at the lower right of the message box.

    The submitted question appears in the conversation above a Processing indicator while the model handles the request.

  9. Confirm that the conversation shows the model's completed answer beneath your question.

    The answer should identify four, although its wording can vary. The response shows local-chat and generation statistics, and the message box is available for another prompt.

Share this guide