AUTONOMOUS DEVELOPMENTSHIP ISSUES,
NOT PROMPTS.
Tag GitHub issues with @taskship. Autonomous coding agents on a workstation we run for you build the branch (frontend, backend, database) and you press Open to load the running app. No chat window. Ever.
HOW AUTONOMOUS CODING AGENTS SHIP A FEATURE
FOUR STOPS. NO PROMPT ENGINEERING BETWEEN THEM.
Tag
Drop @taskship on any GitHub issue, or trigger a shipment manually from the dashboard.
Bundle
Related issues and their milestone become one shipment: a single execution with a manifest.
Build
An agent on a workstation of its own runs your full stack on a branch, however your taskshipyard/ folder starts it (npm run dev, sst dev, docker compose).
Open
Press Open: the live branch app loads from the dev machine in a new tab. Review, then merge.
AI AGENT PREVIEW ENVIRONMENTS, NOT PATCH FILES
EVERY RUN ENDS ON AN APP YOU CAN CLICK.
The machine that built the feature keeps serving it: the frontend on the branch, the API on its own cloud stage, a database with the run's migrations applied. Reviewing is clicking through the running app, not reading a hundred-file diff. How preview environments work →
YOUR ARCHITECTURE,
BEFORE & AFTER
Every shipment changes your system. The Architecture deck tracks the live dev branch and shows, for every shipment, exactly what changed: every lambda, table, and module the agent touched.
BRING YOUR STACK.
CARGO SPEC v1A taskshipyard/ folder in your repo says how your dev environment boots (a shell script or docker compose), and the agent runs and screenshots it from there, in a sandboxed workstation we run for you: one machine per customer, never shared. Connect an AWS role and the cloud stage it deploys lands in your own account. Any framework that serves a page works. SST + AWS + Supabase + React is the most battle-tested path: the stack we run on.