Deep Work Plan
Open-source framework turning repos into execution harnesses for agents.
Visit website →About
Deep Work Plan is an open-source framework that turns a Git repository into an execution harness for AI agents. Rather than dropping context-heavy prompts into a chat window, you embed atomic tasks, acceptance criteria, and validation gates directly in the codebase as structured plan files.
Agents read those plans, execute one task at a time, and check their own work against the validation gates before moving on. The result: agents stay focused on long-horizon work spanning days, and they survive context resets without losing track of the overall goal.
Released under the MIT license by Sergio Florez and collaborators, Deep Work Plan is provider-agnostic — it works with Claude, GPT, Gemini, or any agent framework that can read a repository.