No lock-in
Not another vendor’s assistant
Jira, Monday, and Linear sell you their AI. ScrumDo runs the agent you already chose: switch Claude for Codex on the same card, no migration, no lock-in.
Bring your own agent
Run the agents you already pay for, on the work, under control.
Claude, Codex, Grok, GitHub Copilot, Cursor: orchestrated from one governed card. A human accepts the spec and approves the plan, the run returns as a real pull request with passing checks, and for agents you run on your own keys, your agent vendor bills the model. Optional AI features use the provider you connect; ScrumDo uses AI for support under its own provider.
run #82 · claude-code · local
✓ completedClaude Code$0.15
Replaced the scaffolded counter with a working Tip Calculator: analyze clean, all 5 tests pass, committed and pushed.
View pull requestbehavior verification · CI6 / 7
spec drift0 / 16
agents you run on your own keys: spend billed by your vendor, $0 markup from us
Governed, not bolted on
Two human gates, one audit trail. Then any agent can run.
A person accepts the sensed spec and approves the plan before an agent acts. The run comes back to the same card as a verifiable result: a real PR, CI behavior checks, drift against the accepted spec, and you can swap in a different agent on the same card.
activity · governed lifecycle
- humanaccepted the sensed spec
- agentposted an implementation plan
- humanapproved the plan
- agentexecuted the approved plan
- agentcompleted the run, PR returned to the card
Why it's different
Why this beats a bolted-on assistant.
Governed
Approval before action, proof after
Every run is gated by a human-accepted spec and an approved plan, and returns as a checkable PR. Not a side-chat. An audit trail.
Zero markup
No markup on agents you run
Other tools mark up AI in the seat price. For agents you run on your own keys, your model spend is billed by your own vendor and ScrumDo charges $0 markup. Optional AI features use the provider you connect; ScrumDo uses AI for support under its own provider.
Proof
Honest capabilities, on the record.
Works with the agents you already run
ClaudeCodexGrokGitHub CopilotCursor
It doesn't stop there.
Customer stories
Draft from the stories behind the work.
The same governed record keeps self-interpreted customer stories and team judgment attached to the card, before any agent touches it.
See how →Flow economics
Prioritize by what delay actually costs.
On Business, Cost of Delay and WSJF make “which release first?” a defensible number, sitting on the same card as the work and the agent runs that deliver it.
See how →Put the agent you already pay for to work.
Connect Claude, Codex, Grok, Copilot, or Cursor to a governed card, keep two human gates and one audit trail, and for agents you run on your own keys, let your own vendor bill the model with $0 markup from us. Optional AI features use the provider you connect; ScrumDo uses AI for support under its own provider.