This is a side project of mine. I have made it as an open source spec management platform. Self host of you want and I run a hosted version.
And a spec is a md (markdown) file within your projects.
As we are building with AI agents either via Cursor, VScode or Antigravity, we can see md files or specs (software term for md files) are fast becoming AI infrastructure. A communication layer.
I have used specs mainly for, - Developing spec for a new feature or a bug, generally a new development artifact and discuss the specs with the technical teams. AI agents are good at exploring the features in technical depth, costs, security aspects etc. that can be in one document or more. If there are changes, we can ask the AI agents to change them. - If we need an analysis from a codebase that we need like a compliance analysis on a certain part. Spec that is filtered for the understanding of the target audience or a tool. I have built cost simulation tools based on the codebase, where I built yaml based infrastructure specs that fed into simulation tool. Beauty of these solutions are, the target audience doesn’t need access to whole codebase. I think spec driven engineering is enabled only by AI agentic era. It was there before but it was time consuming and not scalable. In essence, - Specs give AI agents the context they need to build correctly. - They also give teams clarity into why and how something was implemented. - Specs are filtered to provide as inputs to other tools or analysis
But AI agentic development is really fast. As with all fast things, we could see two problems were emerging:
- Discussions across teams became long and fragmented. Compliance needs those specs. Design needs those specs. Risk Modeling needs those specs. - Specs were duplicated, scattered, and hard to reuse across projects.
I built mdspec to remove that friction. I made it open source to make sure anyone can bend it any way they see fit for their own use cases.
mdspec is a specification management platform designed for fast phased AI-driven development. It helps you: - Create structured, reusable specs - Link specs across features and projects - Maintain clarity as development speed increases - Give AI agents reliable, durable context - When specs become first-class citizens, teams move faster, without losing alignment. - Go cross team securely.
I'm exploring the possible integrations to integrate this back into AI agents and external tools.
No comments yet.
