Crypts and Commits: A New Framework for AI Agent Teams
Discover Crypts and Commits, a new development tool blending tabletop gaming metaphors with AI coding agents to enhance human-AI collaboration.
Stock photo for illustration only, not from the actual event
- Crypts and Commits introduces a structured framework for teams of developers working alongside AI agents.
- Inspired by Beads and Gas Town, it uses tabletop gaming concepts as its core architectural metaphor.
- The human developer acts as the Game Master while AI agents play the roles within a defined project world.
- It strikes a balance between reusable package structures and bespoke, project-specific configuration.
While the current tech discourse frequently highlights a single developer coordinating forty coding agents, many development organizations face a much more immediate and grounded challenge: a team of ten engineers, each collaborating with one or two agents, trying desperately to stay synchronized without falling back into the traditional trap of the hero coder bottleneck.
This very organizational problem inspired the creation of Crypts and Commits. The project draws its primary technical influence from Beads, utilizing a git-backed record of work. The creator realized that documentation has evolved from messages left for human colleagues into durable context that an AI coding assistant can directly consume, capturing the deep technical rationale behind changes rather than competing with project management boards like Jira.
Stock photo for illustration only, not from the actual event
Another major influence came from the strategic use of metaphors, much like the Gas Town model. By borrowing vocabulary from outside standard software engineering, systems can drastically shorten conversational loops with AI models. Crypts and Commits leverages tabletop gaming terminology: the human takes the role of the Game Master, and the agents act as the players, establishing a coherent world and shared vocabulary.
Applying tabletop role-playing frameworks to agentic workflows is an innovative approach to mitigating 'Prompt Drift' and context loss in large codebases. Defining explicit rules, boundaries, and character roles provides both human engineers and LLMs with a shared mental model, significantly improving task verification and system exploration.
This design philosophy strongly aligns with Steve Yegge's recent essay, 'The Shape of Things to Come,' which argues that AI harnesses cannot simply be generic, drop-in frameworks. Instead, they must be chemically bonded to the specific application they serve, requiring engineers to explicitly define boundaries, configurations, and system behaviors.
Consequently, Crypts and Commits positions itself in a middle ground. While the package provides reusable structures for documentation, encounters, and repository-backed receipts, the actual execution remains entirely bespoke. Developers must define their own project worlds and decide how agents operate within them.
Ultimately, this workflow is not about delegating a task ticket and walking away. Instead, the human and the AI act as reasoning partners, where the AI generates structured encounters containing requirements and rationale, allowing the human developer to refine the plan until it reaches the desired end state.
Source: Dev.to
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment