AgentsMedium impactFor DevGitHub AI Agents · June 3, 2026
AI-native publishing system - build books like software. 6-phase pipeline from blank page to EPUB/PDF, with 16 skills, 7 agents, and 8 genre templates. Claude Code plugin.
epicsagas/Velith
Velith is an AI-native publishing system that uses multiple AI agents and skills to produce books from blank pages to EPUB/PDF formats, integrating with Claude Code.
Signal strength3.9/5·8 stars
Velith is an AI-native publishing system that uses multiple AI agents and skills to produce books from blank pages to EPUB/PDF formats, integrating with Claude Code.
TL;DR
Velith is an AI-native publishing system that uses multiple AI agents and skills to produce books from blank pages to EPUB/PDF formats, integrating with Claude Code.
What happened
The Velith system has been released as an open-source AI agent framework for automated book creation using a structured 6-phase pipeline, featuring 16 specialized skills, 7 agents, and genre templates.
Why it matters
It demonstrates a novel approach to content generation by orchestrating AI agents for creative publishing workflows, potentially streamlining and automating book production leveraging LLMs.
Generating deep dive...
AI-powered analysis takes a few seconds
The bigger picture
Velith exemplifies the movement from monolithic AI content models towards finely coordinated agent frameworks that simulate software project management within creative domains. This signals a broader industry shift where AI is no longer just a text generator but a component in complex workflows involving planning, revision, formatting, and genre-specific constraints. By mimicking software engineering practices, the publishing pipeline reduces manual overhead and improves output customization, foreshadowing how AI will embed into established media production processes. The availability of open-source, agent-driven frameworks also democratizes access to sophisticated automation that was previously siloed in proprietary platforms. Ultimately, Velith’s approach highlights a trajectory where AI augments authorship through systems thinking rather than isolated generation tasks.
Technical deep dive
Velith’s architecture orchestrates seven discrete AI agents, each presumably tasked with a specific phase such as ideation, drafting, editing, fact-checking, formatting, and export. The 16 skills represent specialized subroutines or capabilities-likely fine-tuned prompts or function calls-that agents invoke based on their phase responsibilities. Coordinating these agents involves pipeline state management to ensure data consistency and context preservation across phases, which is pivotal when transforming raw text into fully structured EPUB or PDF files. Genre templates encode domain-specific rules, narrative structures, and stylistic conventions that agents reference, ensuring output coherence. The integration with Claude Code enhances flexibility by allowing code-driven control and debugging, which provides a robust interface for iterative development and extension. From an engineering perspective, Velith introduces significant challenges around inter-agent communication protocols, prompt engineering modularization, and error recovery in multi-step generation. Developers adopting Velith need to focus on robust versioning of pipeline stages, monitoring agent outputs for drift, and fine-tuning both agents and skills for genre-specific fidelity.
Real-world applications
1
Automated creation of genre-specific novels or novellas, where writers input minimal prompts and receive formatted EPUBs ready for digital distribution.
2
Educational publishing workflows that generate tailored textbooks or study guides by orchestrating agents to research, write, and format content for specific curricula.
3
Self-publishing platforms integrating Velith to offer authors AI-assisted manuscript generation along with instant formatting and export tools for multiple book formats.
4
Creative writing workshops or content studios using Velith to prototype story drafts, iteratively revised through multi-agent feedback loops before manual refinement.
What to do now
Review the Velith GitHub repository and experiment with its 6-phase pipeline to understand integration patterns among agents and skills.
Develop custom genre templates or additional skills to extend Velith’s applicability to niche content domains or vertical markets.
Integrate Velith’s pipeline into existing publishing or content management systems to assess automation impact on workflow efficiency.
Contribute to the open-source codebase by improving agent communication protocols or designing robust quality assurance tests for multi-agent consistency.