The software stack

A stack for working with Prism files

The libraries are how you build on Prism. The software is how you work with Prism files directly — three programs that share one document model and cover the three ways data gets touched: in scripts, by agents, and by hand. Each has its own docs:

• Prism CLI → — the command line: validate documents, convert between formats through Prism, and pack a directory tree into one file. The automation surface.
• Prism Server → — a bridge that lets external tools, including MCP clients like Claude Code, operate on Prism files through stable typed operations. The agentic surface.
• Prism Editor → — a native editor with the server embedded, so you watch an agent's edits land live. The human surface.

One document, three doors

The point of the stack is not three separate tools — it is that they are the same document model, reached three ways. A .prisma file you prism check on the command line is the same Document the editor opens and the same Document the server hands an agent. An edit is an edit regardless of origin: the editor's panels, a prism convert, and an agent's set_property call all go through one typed mutation surface, validated and undoable the same way.

prism convert draft.md draft.prisma     # bring existing content in
prism-server draft.prisma               # let an agent expand it…
prism-editor draft.prisma               # …and watch, or edit alongside it
prism pack ./project project.prism      # ship the whole thing as one file

That is what makes the workflow feel seamless rather than integrated-after-the-fact: at every step it is the same file, never losing fidelity. Pick a tool above for its full documentation.