The project's memory for AI-speed development
Tasks, their history, evidence and decisions kept in one place, distilled by Eve and read by coding agents token-efficiently. Run a project in Vitrinka or mirror it with Jira.
Start free01
A memory, not a tracker
Jira and Plane were designed when a ticket was written once by a person and read a few times. Agents now write and read hundreds of times a day. Vitrinka keeps every change as a ledger row, every upload as a version and every transcript, decision and commit attached to the task by kind. Eight task types, custom fields, labels, sprints and cross-project links cover the planning; nothing is replaced along the way.
Read the guidePM, in depth02
An engine that agents can read
Rules fire from the ledger on events or a schedule, and intake collects proposals from sessions, reports, pull requests and Eve as drafts that stay invisible until someone accepts them. An agent asks for a report and the server counts in SQLite instead of paging through tasks; Eve distils the references into the morning brief and proposes plan changes that a person still has to accept. The engine works with no AI backend at all.
Read the guidePM, in depth03
Keep Jira where the team needs it
The Jira mirror is a per-project choice with a chosen truth side and a state map. A project can run in Vitrinka on its own, or connect to Jira and keep working in both. Commits reach the task through a trailer the CLI writes, and a merged pull request can complete it. Review the mapping and conflict rules before enabling a mirror; nothing is moved out of Vitrinka and nothing depends on the tracker.
Read the guidePM, in depth