Give the agents you use the context they are missing.
Claude, Cursor, Codex and Copilot can read your code. They can't see your revenue, your goals or what is next on the roadmap until you connect them to Shipchart. That connection is most of what Shipchart does with AI. There are also a few built-in tools for thinking a plan through.
Goals: 3 on track, 1 at risk
Stale: cli-tool, no check-in for 23 days
Connect an MCP client
and ask about your portfolio.
Shipchart runs an MCP server. Point Claude, Cursor or any other MCP client at it and the assistant can look up your projects, revenue, expenses, goals, roadmap and check-ins while you talk to it.
- ›Claude.ai and Claude Desktop connect with OAuth. Claude Code, Cursor, Windsurf and others use an API token.
- ›On the free plan the assistant can read everything. On the Builder plan it can also log a check-in, record revenue or add a roadmap item.
- ›The shipchart:init prompt reads the codebase you are in and fills in the project's tech stack and conventions for you.
Or export the context
from a terminal.
If your agent doesn't speak MCP, the CLI does the same job. It prints a project's context as markdown, which you can paste into a conversation or write into the repo for the agent to read.
- ›shipchart context export prints a markdown summary in small, medium or large.
- ›shipchart context pull writes it to AGENTS.md, and to CLAUDE.md as well if you ask.
- ›Any command takes --format llm for compact output meant for a model to read.
When you want to hand
the work off.
On the Builder plan you can assign a GitHub issue to Copilot, pass a Linear issue to a Linear agent, or run shipchart worker so Claude Code, Codex or Gemini does the task on your own machine and opens a pull request. Shipchart keeps a list of what you handed off and what came back.
How delegation works ›Email me when this month's AWS bill is running well above last month's, with the services that caused it.
Compare each service against its own baseline to cut false alarms.
AWS cost data arrives a day or so late, so same-day alerts will misfire.
One email a day, or an alert as soon as a threshold is crossed?
Run the comparison in a queued job and cache the cost lookups.
A few AI tools
for thinking a plan through.
The other half is smaller. Think With Me is the one you will notice first. Switch it on while you write a roadmap item or an idea, and it adds short notes beside your text under four headings: ideas, warnings, decisions and technical points.
It knows the project's stage and tech stack, it can lean toward product or technical advice, and it says nothing when it has nothing useful to add.
Challenge Me
Reads a roadmap item and its work items, then lists scope concerns, things you have missed, feasibility risks and questions to answer first.
Work item generation
Breaks a roadmap item into three to six work items with acceptance criteria. It can ask a few clarifying questions first if you want it to.
Check-in drafts
Writes a draft check-in from the GitHub or Linear issues you closed since the last one, and suggests a mood. You edit it before saving.
Error to issue
Turns a Sentry or Bugsnag error into an issue with a likely cause and a suggested approach, and keeps the link back to the original error.
There is a name generator too, for the projects that are still called untitled-3. These tools spend AI credits, where one credit is roughly one cent of model usage, and the Builder plan includes a monthly allowance.
See each tool in detail › Start with the dashboard,
then connect your agents.
Free for up to 4 projects, with read-only access from MCP, the API and the CLI. The Builder plan adds write access, delegation and a monthly AI credit allowance.