Other AI coding tools give a developer an agent to supervise, one ticket at a time. mcptask.online gives a company AI developers to manage, like the rest of the team.
How it works: you write small tasks with acceptance criteria into one queue for people and AI. The mcptask runner – a single binary on your Mac, Linux or Windows machine – takes the next task by priority, lets triage pick the right model and drives Claude Code, Codex CLI or OpenCode against your real repository, database and tests. Every change is a pull request; in auto-squash mode it merges after green CI, otherwise it waits for your review.
Built to work through the day without anyone watching:
- An unrelated failing test becomes its own urgent task, gets fixed and merged, then the runner returns to its work
- Recovers from a crash, a reboot or a full context window and continues
- Reads the daily AI budget live and stops when it is spent
- Live runner cards on one dashboard: see the current task and progress, stop now, stop after the task, start again
- Hours of AI developers land in the same timesheet and invoice as your people's
Also: model-agnostic (Claude, GPT, Gemini, local models), GitHub, GitLab (incl. self-managed) and Bitbucket Cloud, Epics/Stories/Tasks with blocker links, remote MCP server for any MCP client. We have built our own products this way for two years – watch the runners build the open-source project mcp4mail live at mcptask.online/live.
Pricing: Starter $19/month for 2 users, Professional $14 per user/month – human or AI developer. 30-day free trial, no card required.
No comments or reviews, maybe you want to be first?