EvoForge
EvoForge is a local-first, desktop 'personal AGI' product for building software, automating workflows, and generating media with AI agents. The product is based on an original agent kernel written in Rust rather than a command-line wrapper. It includes 98 native host tools and 54 original skills, allowing agents to interact with the shell, file system, browser, desktop applications, documents, terminal output, email, and multimodal media backends. Users can connect either local models or cloud frontier models, and data and model keys can stay on the user's machine. For software developers, EvoForge supports frontier-level coding workflows. The mutation gate runs a change in an isolated worktree and requires the regression tests to first fail (red) and then pass (green) before code is merged, adding guard tests that accumulate over time. The system can coordinate work through a 'lead' mode—delegating subtasks—or a 'swarm' mode, where multiple specialist agents claim tasks from a shared board with leases and peer notes. EvoForge handles every modality in and out. It can read text, images, audio, video, and documents in all common formats; it can produce code, essays, slides, documents (docx/pptx/xlsx), images, voice, music, podcast episodes, video clips, and finished films. The director's console offers a multi-track timeline for film creation: narration becomes the master clock, placeholders are filled by generated or recorded assets, and a human reviews and exports final media. Free to download, EvoForge targets Windows and macOS and provides six interface languages. While it is positioned as a personal AGI, practical use may require sufficient local hardware or API keys for advanced multimodal generation. The code base is not published; the primary source of information is the official homepage.
- Category
- AI Agent
- Pricing
- Free
Problem solved
EvoForge addresses the fragmentation of AI tasks by providing a local desktop environment that unifies coding assistance, workflow automation, multimodal generation, and multi-agent collaboration. It lets individuals build and run a personal AGI while keeping data, tools, and keys on their own machine, and enables users to direct agents through known control flows rather than relying on wrappers.
Use cases
- Autonomously implement features or fix bugs using a mutation gate that requires failing and passing regression tests.
- Coordinate a lead or swarm of AI agents to tackle a large coding project across a codebase.
- Convert documents such as PDF, PPTX, DOCX, and XLSX into structured text or generate polished output files.
- Create narrated films and videos via the director's console, using generated or recorded shots, music, and subtitles.
- Generate and edit images, including icons and posters, while preserving original context through path-referenced transcript images.
- Transcribe, understand, or generate speech and music using multiple providers or local backends.
- Automate repetitive desktop, browser, and shell tasks by combining 98 native tools with 54 skills and plugins.
Audience
- Developers
- Content Creators
- ML Engineers
- Researchers
- Hobbyists
Pros
- Local-first by design: runs fully on your computer and keeps model keys and data on-device.
- Original Rust kernel, not a wrapper, gives direct control over tools, files, browser, shell, and desktop.
- Broad native multimodal support: ingests and creates text, images, audio, video, and office documents.
- Reliable code workflow with mutation gate and red/green regression tests protecting every change.
- Flexible agent orchestration modes (lead vs swarm) and the ability to grow new skills and plugins.
Cons
- Only provided as free binary for Windows and macOS; no Linux version or source code.
- May need a powerful computer or external model APIs to fully use local models and video/music generation.
- Proprietary desktop app; not open source and no independent code auditing.
- Ecosystem still young; little third-party documentation or community support compared with long-established tools.
- Feature breadth makes onboarding complex for nontechnical users despite a polished interface.
Official website: https://evoforge.cc/
Canonical LambdaBase page: https://www.lambdabase.com/tools/evoforge