memfork 0.3.0

Git for agent memory, shared by every AI tool on your machine: fork, merge, rewind and hand off work between agents over MCP.
Documentation
description = "Ship a feature: agree what it does, build and test it, document it, check it end to end"

# Fill in each empty `accept` once with a command that exits 0 when that task
# is done, such as your test command. An empty one is never run.

[[task]]
id = "design"
title = "Agree what the feature does, including its edge cases"
detail = "Record the behaviour and the reason for it as a decision, so every agent builds the same thing."

[[task]]
id = "build"
title = "Build the smallest change that delivers it"
detail = "Fork memory before anything risky, and discard with a lesson if an approach does not work."
depends_on = ["design"]

[[task]]
id = "tests"
title = "Write tests for the behaviour and each edge case"
depends_on = ["design"]
accept = ""

[[task]]
id = "docs"
title = "Document it where its users will look"
depends_on = ["build"]

[[task]]
id = "verify"
title = "Run everything and try the feature end to end"
depends_on = ["build", "tests"]
accept = ""