Our take
Activeregular commits and releases- Activity
- 79
- Community
- 35
- Issues
- 58
- Pull requests
- 83
+56 stars in the last 30 days
- Merges more pull requests than 93% of the projects we track
Guaardvark is an unusually broad local-AI studio with a fast release cadence, but its scope and strong reliance on one contributor make adoption best suited to teams prepared to operate a substantial self-hosted stack.
Good fit if
- You want local chat, RAG, agents and media generation in one GPU-aware application.
- You can run the Linux-first stack and have hardware appropriate to the workloads you need.
- MCP access, human approval gates and MIT licensing suit your integration and deployment needs.
Look elsewhere if
- You need a small, single-purpose tool rather than an application spanning agents, RAG and media production.
- You need Windows support that is already verified, or video generation on hardware below the documented GPU requirements.
- Your adoption policy requires development to be broadly distributed across contributors: despite 17 contributors, recent work is heavily concentrated in one person.
All health signals
| Last commit | 2026-10-17 (1 day ago) |
|---|---|
| Commits, last 90 days | 100+ |
| Releases, last 12 months | 20 (latest v3.2.1, 2026-10-12) |
| Contributors | 17 (top contributor: 90% of commits) |
| Issues closed, last 90 days | 3+ (typically closed in 2 days) |
| Pull requests merged, last 90 days | 90 (typically merged in 0 days) |
| Project age | 7 months |
Checked on 2026-10-19 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
320Forks
61Change since Nov 13, 2026: +2 stars·0 forks
Overview
Guaardvark is a self-hosted AI studio for running chat, retrieval-augmented generation (RAG), agents, coding swarms and media workflows on your own GPU. Its local stack uses Ollama for language models and ComfyUI for image and video workflows, alongside Guaardvark's own GPU services. The project is aimed at people who want one install to handle these tasks locally rather than rely on hosted AI services.
Linux with an NVIDIA GPU is the primary platform. Apple Silicon is supported with some GPU features using Metal, while Windows through WSL2 is still being verified. The project emphasizes local operation and human approval gates for actions such as publishing and code changes.
Key Features
- Chat with local models and ask questions about indexed files, with hybrid keyword and vector retrieval, reranking and citations.
- Run screen agents in an isolated Ubuntu desktop, with a visible browser and streamed reasoning.
- Launch parallel coding agents in separate Git worktrees, using Claude Code or local Ollama-based backends.
- Connect coding assistants through an MCP server, agent skills and a Claude Code plugin.
- Generate images, video, music and speech locally, with support for LoRA training and imported LoRAs.
- Produce videos through a multi-role Film Crew workflow or a beat-synced music-video director.
- Coordinate GPU services, jobs, plugins and multi-machine nodes through a shared orchestrator.
Use Cases
- A local-AI user can index project documents and query them with cited passages instead of sending files to a hosted service.
- A creator with a capable GPU can generate and assemble images, video, music and voice in one studio workflow.
- A developer can delegate a refactor or test expansion to parallel agents working in isolated Git worktrees.
- A Claude Code, Cursor or other MCP client user can invoke Guaardvark's local tools from their coding assistant.
What you need
Detected in the repository
- Python (main language on GitHub)
- A Compose file to run it with its related services
- Automated checks on GitHub Actions
License in plain words
MITpermissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the copyright and license notice
- Share your changes: no
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
On Linux with an NVIDIA GPU, install and start the application with:
curl -fsSL https://guaardvark.com/install.sh | bash
Then open http://localhost:5173. Review the README and installation guide for hardware tiers, manual setup and other installation options.
Alternatives
- MaxKB: MaxKB focuses on knowledge-backed agents and workflows, while guaardvark bundles local-GPU chat, coding, and media-generation tools.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| guaardvark | Python | MIT | 270 | Active |
| MaxKB | Python | GPL-3.0 | 22.9k | Active |
Considerations
- The full stack has significant hardware and service requirements. The README specifies Python 3.12, Node.js, PostgreSQL, Redis and Ollama; video generation generally calls for a 16 GB-class GPU, with a lower documented threshold for the Wan 2.2 5B model.
- Linux with NVIDIA is the primary target. Apple Silicon support is partial, and Windows through WSL2 is still being verified.
- Development is active, with at least 100 commits in the last 90 days and 20 releases in the last year, but about 90% of contributor activity comes from one person. That concentration is a maintenance risk for a project with a wide feature surface.
- The README lists a hand-editable video-editor timeline as planned work, and the repository metadata does not indicate a test suite, despite having CI.
- The project is MIT-licensed. The README notes that the Guaardvark name and logo are trademarks and are not covered by the code license.
Found this useful?
Share it with someone who would like guaardvark.