Loading...
Product
Core Features
Top Extensions
Company
Use Ollama models from Raycast for local and remote inference, chat, writing assistance, image understanding, model management, and tool calling.
This extension is not affiliated with Ollama.
You can install a model with the Ollama CLI or pull one directly from the Manage Models command. Browse available models in the Ollama library.
Remote servers can be configured without authentication, with basic authentication, or with a bearer token.
Use Chat with Ollama for saved, multi-turn conversations. You can choose separate models for general chat, vision, and tool use, as well as configure thinking effort and how long each model stays loaded in memory.
The action menu lets you:
⌘ M on macOS or Ctrl M on Windows.⌘ T on macOS or Ctrl T on Windows.⌘ B on macOS or Ctrl B on Windows. This requires the Raycast browser extension.⌘ I on macOS or Ctrl I on Windows. This requires a vision-capable model.The Chat Memory Messages preference controls how many recent messages are sent back to the model. The default is 20.
The extension includes focused commands for common tasks:
Most text commands use the input source selected in the extension preferences: Selected Text or Clipboard. You can also enable fallback to the other source when the preferred source is empty. After a one-shot command finishes, use Continue as Chat to keep the conversation going.
Commands that rewrite their input also offer Auto-Replace Selected Text, enabled per command in its Raycast settings. When enabled, the generated text replaces your selection as soon as it is ready, instead of waiting for you to copy or paste it. It only replaces text when the selection actually supplied the input: if the input source was the clipboard, or fell back to it because nothing was selected, the answer is shown instead of overwriting whatever happens to be highlighted. Commands that explain, describe, or summarise their input do not offer it, because their output is not a replacement for what you selected.
Use Manage Models to:
Use Loaded Models to see the models currently held in memory and unload them when they are no longer needed.
Use Create Custom Command to create a Raycast Quicklink backed by your own prompt, model, and Ollama parameters. Prompts follow the Raycast Prompt Explorer format and support these tags:
{selection} inserts selected or clipboard text.{browser-tab} inserts the current browser tab as Markdown. Use {browser-tab format="html"} or {browser-tab format="text"} to request another format.{image} attaches a JPEG or PNG image from Finder or the clipboard. This requires a vision-capable model.Tick Auto-Replace to have the result replace the selected text automatically, as the built-in commands can. This only makes sense for a prompt that uses {selection}. The choice is stored in the Quicklink when it is created, so changing it later means creating the command again.
The extension exposes Web Search and Web Fetch tools to Raycast AI and can also use them inside Chat with Ollama when internet search is enabled.
These tools require an Ollama API key. Create one in your Ollama account settings, then add it to the extension's Ollama API Key preference. An API key is not required for local inference.
Use Manage MCP Server to add, edit, and remove stdio-based MCP servers for Chat with Ollama. Paste a standard JSON configuration such as:
{
"mcpServers": {
"example": {
"command": "npx",
"args": ["-y", "<mcp-server-package>"]
}
}
}
Select the configured servers in the chat model settings. Only MCP tools are currently supported, and the selected Ollama model must support tool calling.
Auto-Replace Selected Text is not listed here: it is set per command, in the settings of each command that rewrites its input.
Chat with Ollama
Summarize content from current Website.
Make selected text more casual
Explain the selected code step by step
Make selected text more confident
Explain selected text in simple terms
Fix selected text from spelling and grammar error
Make selected text more friendly
Describe content of the image on the clipboard or selected from finder
Get text from image on the clipboard or selected from finder
Improve writing of selected text
Make selected text longer
Make selected text more professional
Make selected text shorter
Translate selected text
Rephrase selected text as Tweet
Manage installed models
Show models loaded on ram
Create a custom command
Create a custom command using a quicklinks
Manage configured MCP Server
Change model, reasoning effort, and keep alive for a specific command
Configure model, thinking effort, and keep alive for each command
Configure global default model settings for all commands
textThinkingBuffer on stream termination (emitDone) combined with a else if in the throttle path that prevented symmetric flushing of thinking and content buffers.keep alive, error missing unit in duration on Command 'Create Custom Command'.keep alive, error missing unit in duration: unit is now required for negative or zero duration time. Use -1s for persistent model on memory or 0s for free up memory immediately.Chat with Ollama.Chat with Ollama and Describe Content of Image.Cannot read properties of undefined (reading 'split') caused by API route '/api/show' responding with empty 'parameters' value for some models.ollama-models | TypeError: Cannot read properties of undefined (reading 'split') when model show information was undefined.ollama-chat | SyntaxError: Expected ',' or '}' after property value in JSON at position 97 when Ollama chat api response with an unformatted JSON response.ollama-models | SyntaxError: Unexpected non-whitespace character after JSON at position 30 when Ollama pull api response with an unformatted JSON response./image permit to add an image from clipboard or more than one from Finder selection on the prompt. At least Ollama v0.1.15 and one multimodal model installed are required. Only PNG or JPG image are supported.CMD+Y to toggle on and off.ModelsOnRegistry.lengh undefined.SyntaxError: Unexpected end of JSON input caused by Ollama Generate API response no longer providing sample_count and sample_duration fields.