How it works
A writing loop built around your judgment.
Calliope treats an AI agent as a capable collaborator inside a project—not as an invisible autocomplete that owns the final word.
1. Begin with durable source files
A Calliope project is designed around material you can inspect and move. The local-first approach means the manuscript is not merely a view into a vendor-owned database.
2. Ask for a bounded piece of work
An agent may help examine continuity, surface a weak scene objective, propose alternatives, or revise a selected passage. Clear scope makes the work easier to evaluate.
3. Inspect the result
Agent output should be reviewable in context. The goal is not to maximize generated words; it is to help the writer make a better-informed decision.
4. Keep or reverse with confidence
Git-backed history is intended to preserve meaningful checkpoints and changes. You should be able to compare the manuscript before and after a revision and recover earlier work.
What is available today?
Calliope is in active development and is not yet generally available. Product details on this site describe the design direction and will be updated as the implementation changes.
Read the product factsFollow the build.
Occasional progress notes and an invitation when testing opens.