Start with context
The agent reads priorities, blockers, incidents, active threads, and quick wins through one MCP call.
Agent-native project management
Your agent writes code. Pulsyr keeps the work legible.
A self-hosted backlog built around MCP. Your coding agent can read priorities, create work, follow dependencies, and close what it finishes—without a second system to update by hand.
MIT-licensed code · Postgres-backed · Works with MCP-compatible clients
One source of truth
Stop translating agent work into status updates.
Most backlogs are designed for humans to update after the real work happens. Pulsyr exposes the same structured project context to you and your coding agent.
Keep the graph current
Create items, add typed dependencies, move lifecycle stages, and attach decisions as the implementation evolves.
Close the loop
Completion records a reason, linked commit, events, and newly unblocked work instead of leaving stale tickets behind.
A workflow with guardrails
Your agent can act. It cannot invent the rules.
01
Connect a project
Each token belongs to exactly one project and carries a read or write scope.
02
Read structured context
Backlog state, graph edges, management docs, pendings, and incidents stay queryable.
03
Move through a lifecycle
Transitions are validated. Terminal work requires a reason. Blocking derives from the graph.
04
Review every change
Humans retain the web interface, event history, role controls, and final judgment.
Designed for focused teams
Small enough to run yourself. Structured enough to trust.
MCP tools
Context, backlog, dependencies, threads, incidents, management documents, pendings, and plans.
Token-to-project isolation
An agent token cannot silently write to the wrong project.
UI languages
English and Spanish are available, with an automatic regional default and a manual selector.
Free plan
Connect your first project without configuring infrastructure.
The Free plan includes one project, one teammate, three MCP tokens and 25 MB of documents. No card required.