Skip to content
Onplana Docs

Create a project from ChatGPT or Claude

All plans Member

Describe a piece of work in one sentence. Get back a real project: epics, tasks, estimates, owners, dependencies, and a schedule where each task starts when the one before it finishes.

Not a draft to review. Not a suggestion to copy out. The project exists in Onplana by the time the answer comes back, and the reply carries a link straight to it.

An Onplana plan rendered inline in a chat: the project name and status, chips reading 19 scheduled, 34 percent complete and a date range, then a Gantt chart with green completed bars, an indigo in-progress bar with a partial progress fill, a red overdue bar beside a dotted today marker, diamond milestone markers, and an Open in Onplana link.

Ask for a plan

Section titled “Ask for a plan”

If Onplana is already connected in ChatGPT or Claude, this is the whole thing. Say what the work is and that you want it in Onplana. The word “plan” steers the agent toward building a structure rather than answering a question:

Plan a website redesign in Onplana. We are rebuilding the marketing
site on the new design system, starting in September. Break it into
discovery and build, with dependencies and estimates.

The more you give it, the less it invents. Durations, an order of work, who owns what, a start date: each one it has is one it does not have to guess.

The reply comes back with a link to the project. Open it and the work is there.

Why the dates are the interesting part

Section titled “Why the dates are the interesting part”

Look at the schedule in the screenshot rather than the task list. Each task starts the day the one before it finishes, because the request declared dependencies and Onplana resolved them into an actual chain.

That is the difference between a plan and a list. Most AI tools hand you tasks that all share one deadline, which tells you nothing about what is blocking what, or what slips when a task runs late. Here, move one task and everything downstream moves with it.

It is a real project, not a preview

Section titled “It is a real project, not a preview”

Open it and you get the epics in the sidebar, the tasks under them, and a Gantt with the dependency arrows already drawn. Nothing is flagged as machine-generated and nothing sits in a draft state waiting for approval. Your team can work in it immediately, and you can edit any of it by hand.

Worth knowing

Section titled “Worth knowing”
  • Tasks are assigned to you unless the plan names someone else, so nothing lands on a colleague without you asking for it.
  • You can add to an existing project instead of making a new one. Say which project and the agent extends it, leaving what is already there alone.
  • Dates come from your estimates and dependencies, not from a calendar the agent invented.
  • One request builds up to 100 tasks and 20 epics. A bigger plan is split across several, which happens without you doing anything.
  • Everything is permission-checked and audited. The agent acts as you and can reach nothing you could not reach yourself. See what agents can and cannot do.

Getting better plans

Section titled “Getting better plans”

Onplana publishes a planning skill that teaches an agent to interrogate the brief before it builds: what the deadline really is, who is available, what the acceptance criteria are. Plans that come out of it are noticeably less generic.

You do not have to install anything to use it. Ask for it in the same chat:

Fetch the Onplana planning skill and follow it.

See Install the Onplana agent skills for per-client setup and the download route.

Section titled “Related”