AI that plans from your own sprints, not from guesswork
Roadmap, backlog, sprints and delivery metrics live in one workspace, so an AI suggestion can point at the work behind it. One feature is live today; the rest is marked as planned.
Most planning tools bolt AI onto a tracker that can see your tickets but not the initiative they deliver, or the velocity that would make a date credible.
PlanMagnet keeps all of it together, which so far buys one live feature and a permanent record of every AI call. Everything else here is honestly marked as planned.
Everything below is labelled with where it actually stands in PlanMagnet today — shipped, in progress, or on the roadmap.
AI-native by architecture, not bolted on
Four architectural choices that make AI a first-class part of PlanMagnet rather than a chat window on the side.
One place to reason from
Roadmap, backlog, sprints, blockers and delivery metrics are the same records. Nothing has to be reconciled against an outside tracker first.
AI suggests, it never writes
The live feature only hands back a list. Nothing reaches your backlog until you confirm it, and then it is created the way you would create it.
Every call leaves a record
Each AI request writes one permanent line: model, timing, what it decided, whether it worked. Failed calls are recorded too, not quietly dropped.
It only sees what you can see
AI requests answer to the same permissions as everything else in PlanMagnet. An outside agent driving it over MCP inherits your rights and never more.
From intent to a governed action
Walk the path a request takes. Select any stage to see what happens there and what backs it.
Stage 1 of 5 · Ground
Read your own history first
The numbers a planning call needs are already sitting in your workspace.
Velocity across your last completed sprints, burndown, cumulative flow and cycle time all come from your real work. Blocked-by links are actual connections between items rather than a note in a comment, so a chain can be followed instead of guessed at.
What the AI in PlanMagnet actually does
Filter by delivery status, then open any capability for the detail and what backs it. No capability is listed as shipped without something in the product behind it.
Where the AI shows up
The places AI meets the work in PlanMagnet — and how far each one has actually got.
MCP & outside agents
ShippedPlanMagnet's most complete AI surface today is outside the app: a published MCP server that lets the agent you already work in read and run your planning.
- 155 typed planning tools across 18 areas of PlanMagnet, plus shared platform tools
- An agent inherits exactly the permissions of the account you connect it with
- Stated plainly: 24 delete tools still have no confirm step, so scope that account tightly
See it as a real scenario
Every AI capability above shows up in a concrete PlanMagnet story. Open one to read the full walk-through.
Turn a paragraph into a reviewable backlogThe one AI feature live today: typed, checked work-item suggestions from a description — with no way to write to your backlog.Read the story
A date you can defend, because it shows its inputsTrailing velocity ships today; the planned forecast adds a date range, a confidence signal and the sprints it was based on.Read the story
Be told what is slipping, and whyA reason per flagged item instead of a score — stale, blocked, over-committed — plus a sweep planned to run without being asked.Read the story
Stop the loudest voice winning on a technicalityPlanned: 'SAML login' and 'enterprise sign-on' counted as one demand, with votes merged so one person is not counted twice.Read the story
Every AI call leaves a recordOne permanent line per call — model, timing, decision, outcome — written before the answer reaches you, failures included.Read the story
Let an outside agent run your planningA published MCP server exposing PlanMagnet as typed tools to the agent you already use — with the delete gap stated plainly.Read the story
Four planning decisions, one at a timeThe scorecard across every AI decision PlanMagnet commits to: which one is live today, and which three are still planned.Read the storyAI you can actually let near your data
An AI-native product has to be governable. Here is where PlanMagnet stands on each control — including the parts still being built.
Nothing lands in your backlog unreviewed
The live AI feature has no way to create a work item at all. What you confirm is created through the ordinary flow, with its usual checks and audit trail.
AI answers to the same permissions
AI requests are permission-checked exactly like every other action in PlanMagnet. An agent driving PlanMagnet inherits its account's rights and never more.
Failures are recorded, not swallowed
A model that errors still leaves a line with the error and the timing. A garbled answer is recorded as such and you get a clear error — never an empty success.
The audit trail is not a copy of your work
Only a short non-identifying summary of what was asked is kept, never your raw text. Model keys sit in secure storage and are read at the moment of the call.
Confirmation for destructive actions
Two dozen delete tools are reachable by an agent with no confirm step or dry run today. Permissions still bound it. Gating them is outstanding, and we say so.
Cost visibility and published test sets
Timing and token counts are recorded today; cost is not, and there is no credit cap or published test set. Metering, caps and test cases are planned.
Models
Gemini, OpenAI and Anthropic are all wired into PlanMagnet, each with a documented default model. Name one for a single request and it is honoured exactly; leave it and PlanMagnet works down the list. Bringing your own key is not offered yet.
- Google Gemini
- OpenAI
- Anthropic
The honest answers
One feature is live: turning a written feature request into checked, typed work-item suggestions. Every call is permission-checked and leaves a record. There is no button for it in the app yet — it runs through the API, the CLI or an agent. Forecasting, at-risk flags, summaries and cost limits are planned.
Put PlanMagnet’s AI to work
Roadmap, backlog, sprints and delivery metrics live in one workspace, so an AI suggestion can point at the work behind it. One feature is live today; the rest is marked as planned.