A few AI tools for planning
what to build next.
Shipchart is a dashboard first. Inside it there are a handful of AI tools that help you think a roadmap item through before you build it, and take some of the typing out of check-ins and bug reports. This page goes through each one.
Think With Me reads along
while you write.
As you type a roadmap description, Think With Me reads it and adds short notes beside your text. It doesn't write the plan for you. It points at the things a colleague would ask about, which is useful when you don't have colleagues. It runs on a small model to keep the credit cost low.
- ›Switch it on while you write a roadmap item or an idea
- ›Notes appear under four headings: ideas, warnings, decisions and technical points
- ›Three modes: Product, Technical and Balanced
- ›It stays quiet when your text is too vague to say anything useful about
Build a billing page with subscription management, plan upgrades, and invoice history...
Consider adding rate limiting before launch. Your API has no throttling and a single bad actor could spike hosting costs.
Your description mentions "user roles" but the schema has no permissions model. This is a scope decision worth making now.
Should the billing page support multiple payment methods or Stripe only at launch?
Stripe webhooks need idempotency keys to avoid duplicate subscription activations.
Will this support multiple payment providers or Stripe only?
Do you need subscription management or one-time payments?
From a roadmap item
to a list of work items.
Write down what you want to build and Shipchart breaks it into implementable steps. You can accept them, edit them or throw them away. If the description leaves things open, ask it to check with you before it generates anything.
- ›Three to six work items per roadmap item, each with a short description
- ›Every work item comes with acceptance criteria you can test against
- ›Tick "Ask clarifying questions" and it checks a few things with you first
- ›It reads the project's stage and tech stack, so the steps fit your setup
Challenge Me picks holes
in the plan.
When you work alone, nobody tells you the plan is too big. Challenge Me reads a roadmap item and its work items and tells you what looks overscoped, what is missing and what will take longer than you think. It is told not to be polite about it.
- ›Four headings: scope concerns, missing considerations, feasibility risks and questions to answer first
- ›Each point is marked critical, important or minor
- ›If the scope looks fine it says so, instead of inventing concerns
- ›Run it before you break the item into work, while changing the plan is still cheap
Multi-tenancy adds 3-4 weeks of work. Consider launching single-tenant first.
No data migration plan mentioned. Existing users will need a path forward.
Real-time sync requires WebSocket infrastructure you do not currently have.
Will free-tier users have access? This affects your conversion funnel significantly.
Work items and roadmap items have a few smaller actions attached, for the moments when a title is all you have written. The name generator sits in the Tools menu on the project page.
Expand
Breaks one work item into three to five implementation steps.
Write description
Turns a short title into a two to four sentence description.
Review
Checks a roadmap item for work you have not listed yet and scores it out of ten.
Name generator
Suggests 8 names for a project, or 15 in extended mode, with a one-line reason for each.
Handle missing subscription on billing page
Add null check for user.subscription before accessing plan details. Redirect to pricing page when subscription is undefined.
Closed six issues since the last check-in, mostly billing: the upgrade flow, invoice history and a webhook retry bug. Suggested mood: good.
Less typing for bug reports
and check-ins.
With Sentry or Bugsnag connected, Error to Issue turns an error into a written issue with a likely cause and a suggested approach. Check-in drafts do something similar for your weekly note, using the issues you closed in GitHub or Linear.
- ›The diagnosis is a best guess and says likely where it is not sure
- ›The issue keeps a link back to the original error, so the full stack trace is one click away
- ›Check-in drafts group the issues you closed into two or three themes and suggest a mood
- ›Both are drafts, and you edit them before anything is saved
Prefer the assistant
you already use?
Everything above runs inside Shipchart. If you would sooner do your thinking in Claude or Cursor, connect the MCP server and the assistant can read the same projects, roadmaps and numbers for itself. Reading is free, and the Builder plan can write too.
You can also save your own agents in Shipchart, each with its own instructions, and export them as SKILL.md files or use them as MCP prompts. Most people won't need to.
How to connect an MCP client ›dashboard_getPortfolio summary with monthly revenue and stale projectsprojects_get_contextTech stack, repo, brief and notes for one projectroadmap_getRoadmap items for one project or allgoals_listGoals across the portfolio, with progresscheckins_createLog a check-in (Builder plan)How AI credits work
Each of these tools spends AI credits, and one credit is roughly one cent of model usage. The Builder plan includes a monthly allowance, your balance is shown in the app, and you can buy more if you run out.
Add your projects, and the tools
are there when you want them.
Free for up to 4 projects. The Builder plan includes the monthly AI credit allowance these tools run on.