Alfrig: An office of AI agents, from request to merged PR

Hi Product Hunt! 👋 I'm Daniel, maker of Alfrig. Every engineering team has a backlog of tasks that are clear, well-scoped and… never the priority: small fixes, utilities, docs, tests, tweaks across repos. I built Alfrig so AI agents can take those off the team's plate, and engineers can focus on the complex, important work only they can do. Alfrig is an office of AI agents that takes a task from request to merged pull request. You hire agents like teammates (name, role, even their look), and they work in a pixel-art office where you can watch what each one is doing. How a task flows: You ask in a chat, or with a label in Jira, Linear or ClickUp, or through the API/MCP An agent investigates the code and writes a spec with acceptance criteria before coding It builds, runs the tests and opens the PR A different agent reviews it, checking every criterion. The author never approves its own work Merge only with green CI. Or stop at any step for a human to approve A few things we care about: 💸 You bring your own model key (Anthropic, OpenAI or z.ai) with no markup. Every task shows what it cost. Our demo task, adding a function with tests, cost US$ 0.02 🔒 The AI never sees your Git token, and each step runs in an isolated, throwaway machine 🧠 Agents save lessons after every task, so the next one starts knowing your project Works with GitHub, GitLab, Bitbucket and Azure DevOps The Free plan is real: 2 agents and 20 tasks a month, no card. Enough to watch one agent build while another reviews. I'd love your honest feedback: which tasks in your backlog would you hand to an agent first, and which steps would you still keep manual? I'll be here all day answering questions! 🐈‍⬛