Wisp: A Hotkey-Driven AI Overlay for Your Desktop

Research context and background

Wisp is an open-source, cross-platform AI assistant designed to enhance your workflow without breaking your flow. Unlike traditional chat applications that require switching contexts, Wisp provides a hotkey-driven overlay that allows you to interact with AI directly within your current application. It reads your selection, clipboard, active app, browser, documents, or screen snippets, streams answers into a compact bubble, and supports voice input and output. Built entirely in Python, it is 100% open-source, extensible, and permissively licensed (MIT).

Benefits

Wisp offers several key advantages for users who want to integrate AI into their daily work without distractions. Its local-first privacy approach ensures that data stays on your machine unless you explicitly choose to send it to a chosen model provider. There is no hosted storage layer, which means your information remains secure. The tool captures specific context like selected text or clipboard content rather than relying on broad screen grabs, making interactions more precise. It is highly customizable with support for addons and the Model Context Protocol, allowing users to tailor the experience to their needs. Additionally, Wisp supports a wide range of LLM providers, giving users flexibility in choosing their preferred AI models.

Use Cases

Wisp can be used in various scenarios where quick AI assistance is needed without leaving your current task. For example, you can highlight text in a document and press a hotkey to get an explanation or rewrite. You can access clipboard content to summarize or translate it. The tool can read the active application window to help with tasks in specific programs. It can also capture open documents and browser pages for analysis. For visual tasks, you can draw a region to send a specific screenshot to a vision model. Recent files can be accessed for quick reference. Voice integration allows you to record and transcribe queries or dictate directly into text fields, making it useful for hands-free workflows.

Pricing

Wisp is free and open-source. It supports free-tier access from providers like OpenRouter, Google AI Studio, Mistral, and GitHub Models. However, free model tiers are rate-limited and subject to change. Users can also configure their own model providers, including self-hosted solutions like Ollama, LM Studio, or vLLM, which may have their own pricing structures.

Vibes

As an open-source project, Wisp has a community-driven approach. The project welcomes bug reports, especially regarding OS permissions, window managers, audio devices, and display servers. Logs can be found in the build_logs folder for troubleshooting. The project is open to contributions via pull requests and donations, indicating a collaborative and supportive community environment.

Additional Information

Wisp is built in Python and is 100% open-source with an MIT license. It supports Windows 10/11, macOS 13+, and Linux (X11 supported; Wayland in progress). The project includes a bundled Model Context Protocol bridge that allows Wisp to use external MCP servers as tools. It also features a local MCP stdio server that trusted clients can use to read live desktop context. The project is actively seeking contributions and feedback from the community to improve its functionality and compatibility across different operating systems.