Set up local models

Worsley captures models you run yourself (Llama, Qwen, Mistral, Gemma, Phi, DeepSeek and others) behind a local model server such as Ollama, LM Studio, llama.cpp or vLLM. A relay sits between your tools and the server, so every tool that talks to it is captured.

Sign in to the Worsley CLI

Do this once on each computer where you run your models.

1. 1Install the Worsley CLI. It runs on macOS, Windows and Linux and needs Node.js 22.23.4 or later (on a Mac, brew install node works too). On macOS or Linux, run:

curl -fsSL https://www.worsley.ai/install.sh | sh

On Windows, in PowerShell:

irm https://www.worsley.ai/install.ps1 | iex

The installer checks the download against its published checksum, puts the CLI in your home folder (~/.worsley, or %LOCALAPPDATA%\Worsley on Windows) and adds it to your PATH. Open a new terminal window afterward so the worsley command is found. Run the same command again any time to update.

1. 1Sign in. The command below prints a link and a short code, then waits.

worsley login

1. 1Open the link. Log in to Worsley if asked, check that the code on the page matches the one in your terminal and that it names this computer, then select Connect. If you last logged in more than 10 minutes ago, Worsley asks you to log in again first. Opened the page without the code? Type it in.
2. 2Back in the terminal, Worsley confirms Signed in as your email. The sign-in is kept in your system's credential store (macOS Keychain, Windows Credential Manager or Secret Service on Linux), never in a plain file. The code expires after 10 minutes; run worsley login again if it does.

Start the Worsley relay

1. 1Start the relay in front of your model server, and leave it running in its own terminal window while you work. Press Ctrl+C to stop it.

Ollama (its usual address, http://127.0.0.1:11434, is the default):

worsley proxy

LM Studio (its local server usually runs on port 1234):

worsley proxy --upstream http://127.0.0.1:1234

llama.cpp's server (usually port 8080), or any other server: pass its address.

worsley proxy --upstream http://127.0.0.1:8080

The relay listens on 127.0.0.1:11435, on this computer only. Port taken? Add --port <number>.

Point your tools at the relay

1. 1Wherever a tool is set to your model server's address, use http://127.0.0.1:11435 instead (add /v1 if the tool expects an OpenAI-style address). Your server keeps running as before; only the address your tools call changes.
2. 2For tools that use Ollama's own settings, set the OLLAMA_HOST environment variable to http://127.0.0.1:11435 before starting them.
3. 3Chat as you normally do. Each exchange is uploaded a few seconds after it finishes.

Check that it's working

• Open History in Worsley. Your session shows up there with the tool's name, usually within a minute of a finished turn.
• In the terminal, run worsley status. It shows the account this computer is signed in to and anything still waiting to upload.

What's captured

• Chats through the OpenAI-compatible chat API, and Ollama's chat and generate APIs: your prompt, the reply, the conversation's earlier turns and the model's name.
• Not captured: system prompts, tool calls and their results, images, and thinking. Everything else your tools send passes through unrecorded.
• Regenerating a reply replaces it in Worsley, so the session follows the version you kept.
• Tokens and timing for every reply: tokens in and out, tokens per second, time to the first token, and how long the model took to load. They're in the session's Performance tab in Worsley, by reply and by model, so you can compare the models you use.
• Switching models part-way through a conversation is fine: each reply keeps the model that wrote it.

If a session doesn't appear

• Browser: check that the Worsley icon isn't showing Not Connected, Reconnect or No Site Access, and reload the tool's tab. More in Extension isn't capturing.
• Terminal: check that worsley watch or worsley proxy was running while you worked, and run worsley status. If it says the sign-in was revoked or expired, run worsley login again.
• Still missing after a few minutes? Contact support and tell us which tool and which browser or terminal you used.