Wolffish
An OS personal desktop AI agent
Last verified:
What is Wolffish?
Wolffish is an open-source, local-first personal AI desktop agent that runs on macOS, Windows, and Linux and gives a user an agent that can both think and take actions on the user's machine. The app maps a modular runtime pipeline (15 runtime modules) to provide deterministic agentic behavior for routing input, building context, enforcing safety, and executing tasks, while keeping data and execution local to the user. It stores the agent's knowledge and configuration as human-readable markdown that you can edit, version-control, and inspect, making the agent's “brain” transparent and auditable. Wolffish targets power users, developers, and privacy-conscious individuals who want an autonomous assistant that automates repetitive workflows, converts and processes media, engages on social platforms, and integrates with local files and tools without sending data to cloud services.
Wolffish pricing
Pricing model: Freemium
Wolffish is presented as an open-source, local-first application with downloadable installers and install scripts for macOS, Windows, and Linux; the website emphasizes local usage and running with the model of your choice rather than listing commercial tiers, so there is no on-site subscription pricing shown and functionality centers on the free, open-source desktop app and self-hosted model usage.
Wolffish pros
- Runs fully locally with no required cloud service
- Open-source codebase for auditability and customization
- Cross-platform support: macOS, Windows, Linux
- Agent brain stored as editable markdown for transparency
- Deterministic modular pipeline with 15 runtime modules
- Full system access to automate local files and apps
- Supports choice of models instead of forcing a hosted model
- Install scripts and installers for each OS for easy setup
- Designed for both thinking and acting (agentic capabilities)
- Built with Electron for a native-like desktop experience
- Use-case guides (media conversion, Reddit engagement, etc.)
- Version-controllable agent configuration and memory
- Safety and routing modules to reduce risky actions
- Local-first approach reduces exposure of sensitive data
- Community and documentation (docs.wolffi.sh) for examples and guides
- Deterministic behaviour suitable for reproducible automation
Wolffish cons
- Requires local compute resources and model management
- Not a fully managed cloud service — user maintains updates
- May be complex for non-technical users to configure
- Some workflows require selecting and hosting your own models
- Electron app size and desktop footprint may be heavy
- Limited mobile or web-first access (desktop-centric)
- Safety relies on local configuration and module setup
- Hardware limitations restrict performance for large models
- Feature set assumes technical familiarity with markdown and version control
Frequently asked questions about Wolffish
How do I install Wolffish on my operating system?
Wolffish provides platform-specific installers and install scripts: a curl-based shell installer for macOS and Linux and a PowerShell command for Windows, plus downloadable installer files (DMG for macOS, EXE for Windows, packaged artifacts for Linux) so you can run the provided install command or download the appropriate installer for your OS.
Does Wolffish send my data to the cloud?
No — Wolffish is local-first and designed to run entirely on your machine so agent execution, context, and memory remain local unless you explicitly configure an external model or service; the default approach emphasizes local data privacy and control.
Can I use my preferred language model with Wolffish?
Yes — Wolffish is built to work with the model of your choice, allowing you to point the agent at local models or external APIs depending on your preferences and resources, rather than forcing a single hosted model.
What kinds of tasks can Wolffish perform?
Wolffish is designed to both think and act: it can automate file and system workflows, convert and process media (media conversion use case), create and publish content (Reddit engagement use case), and execute scripted automation tasks using its modular runtime pipeline and full system access.
How is the agent’s knowledge stored and edited?
The agent’s brain is stored as human-readable markdown files that you can open, edit, and version-control, giving you direct visibility into prompts, memory, and configuration and enabling reproducible changes and audits.
Is Wolffish suitable for non-technical users?
Wolffish targets power users and developers; while installers simplify setup, many of its strengths (choosing models, editing markdown brain files, configuring modules) assume technical familiarity, so non-technical users may face a learning curve.
What safety measures does Wolffish include?
Wolffish’s runtime architecture includes modules for routing input, building context, and enforcing safety; these components are part of the deterministic pipeline intended to gate risky actions, but effective safety depends on local configuration and the models you plug in.
Can I version-control my agent and its configuration?
Yes — because the agent brain and configuration are stored as markdown and files on disk, you can use git or other version-control systems to track changes, revert configurations, and collaborate.
Are there example use-cases or guides available?
The Wolffish project publishes use-case guides and documentation (for media conversion, social engagement, and other workflows) to help users understand and implement common automation patterns with the agent.
What are system requirements for running Wolffish effectively?
System requirements depend on your choice of models and workloads; Wolffish itself runs on macOS, Windows, and Linux, but performance for large or on-device models requires sufficient CPU/GPU and memory resources, while lighter setups can use external model APIs if local hardware is constrained.