Lifecycle-aware backlog
Items move through validated states from idea to done. Closing requires a reason, and every mutation leaves an event trail.
Product
A backlog designed for work done by people and agents.
Pulsyr keeps project state structured, permissioned, and visible. The web interface remains human-friendly while MCP gives coding agents a deliberate way to read and update the same system.
Dependencies that compute
Typed relationships—blocks, requires, conflicts, related, and part of—turn blocking into derived state instead of another checkbox.
Project-scoped agent tokens
Read and write scopes are explicit. A token is bound to one project, so context and actions cannot drift into another workspace.
Priority and search
Filter by lifecycle, area, priority, and effort. Optional AI enrichment can estimate impact and surface quick wins.
Threads and management
Move large ideas through investigation and specification, then track documents, pendings, and a three-level Gantt plan.
Incidents and webhooks
Bring Sentry issues into a triage container and close work from signed GitHub webhook events linked to a commit.
Human control stays visible
Automation without a black box.
The lifecycle rejects illegal moves instead of silently accepting inconsistent state.
Owners, members, viewers, and superadmins see only the controls their role permits.
Without external AI keys, the core backlog continues to work and features degrade gracefully.
You choose the server, database, integrations, retention, network, and upgrade schedule.
Start with your own infrastructure
Use the product and the protocol together.
Start with the Docker image, then connect one project token to your MCP-compatible agent.