What is farmtable?

Farm Table is an open source task runtime built specifically for AI coding and execution agents. It gives those agents one predictable interface for receiving work, understanding how tasks depend on each other, and tracking progress, whether the tasks actually live in GitHub, Jira, Linear, or Farm Table's own built in task store. When no outside platform is connected, that built in store still works fully on its own, tracking dependencies, handling safe concurrent claims, and instantly finding which tasks are ready to work on, with no API keys or accounts needed. The project describes itself as early and experimental. It runs in two modes from the same codebase. Embedded mode is the default: everything runs in a single process with a local SQLite database, so it works with zero extra infrastructure. Client-server mode connects a command line tool or agent to a separate server backed by Postgres, meant for coordinating multiple agents at once in production. Every task, whether synced from an outside platform or created locally, is represented the same way, with status split into a high level phase, a more detailed stage, and the platform's own native label, plus fields for the related Git repo, branch, pull request, and CI status. Tasks form a dependency graph, and Farm Table can list which tasks are ready to start, walk the full tree of what blocks or is blocked by a task, calculate the longest blocking chain in a set of tasks, and point out which tasks are blocking the most downstream work. Running one command turns Farm Table into a Model Context Protocol server, exposing ten tools so any compatible AI client can list, claim, and resolve tasks directly. The project also ships a web dashboard built with Vite, TypeScript, and Lit web components, offering a Kanban board, a visual dependency tree diagram, and an inspector panel for task details, comments, and history. Setup for local development needs Go 1.22 or newer and Node.js 22 or newer, and the project can also be built into a single Docker image for production use with Postgres. It is released under the Apache License 2.0.