The problem
AI agents re-reason through every task from scratch, even ones they’ve solved before:- Inconsistency: Agents take different paths or skip steps between runs — there’s no way to verify without watching
- Cost: Each reasoning pass costs tokens — the same tokens, for the same logic, every time
- Context bloat: Loading tool schemas (especially MCP servers) consumes tokens before any work begins
How pflow helps
pflow separates planning from execution:- Your agent plans once - figures out what nodes to use and how to connect them
- pflow compiles the workflow - saves it as a reusable
.pflow.mdfile - Execution is instant - run the same workflow with different inputs, zero reasoning cost
Workflows are documentation
A pflow workflow is a.pflow.md file — a standard markdown document that happens to be executable. Headings describe inputs and steps. Prose explains intent. Code blocks contain the actual commands, prompts, and configuration.
The format is markdown because that’s what agents naturally produce — headings, YAML, code blocks. No framework to learn, no class hierarchy, no imports.
Open one on GitHub and it reads like documentation. Run it with pflow and it runs as a workflow. Same file, both purposes. No separate documentation to maintain, no opaque config to decipher — the workflow explains itself.