Dispatch the whole backlog at once — every agent on its own git worktree, wired to the tools you already use. You come back to a review queue, not a wall of files. Everything ships as a PR you approve.
Coding agents got good at writing code. What they didn’t bring is a way to run six of them at once, keep them working from the same brief, and hand you back something you can actually judge. That’s the whole of fables.
Dispatch the whole backlog at once. Each ticket gets its own agent and its own git worktree, so nothing collides. You come back to a queue — diff, tests, risk — and merge or send it back in a keystroke.
Plan tickets from Linear, Jira, or GitHub Issues. Dispatch an agent and it spins up an isolated git worktree, implements, writes tests, and opens a PR for you to review.
gh — nothing proprietary in the loop.Sentry crashes, PostHog drop-offs, Vercel build failures — they all surface in one place. One click turns an incident into a tested pull request with the repro baked in.
Edge-to-edge, full-screen when you want it. Real git, gh, npm, and claude — nothing to relearn, nothing flaky to wrap. The agents drive the same shell you do.
Claude Code is native and available today — sign in and sessions, skills, tools and subagents are first class. The rest arrives over the Agent Client Protocol, the open standard between editors and agents, so adding an agent becomes a connection rather than a rewrite.
Every run leaves a trail: the intent it started from, the context it was given, what it changed, what the tests proved, who approved it, and what happened after deploy. Your history of shipping with agents, in one place.
Because every run is recorded, the numbers come for free — what shipped, how long review took, which runs came back. Then fables reads the week and tells you what it means: where the time actually went, what’s slowing you down, who is carrying what.
Most of the agent time went to the refunds epic, and it moved. Review is now the slowest step — PRs waited 9h on average, twice last week. Two runs came back for the same missing e2e cover; worth a skill.
Open a project and you see who else is in it — and where. Attach to a running agent’s terminal and watch the same stream your teammate is watching, then hand the review over without a screen share.
Add official Claude skills — Next.js, Supabase RLS, Sentry triage — and your agents follow your conventions automatically.
Cut a version and publish to npm with provenance, or push a Vercel and Expo build — tracked in the workspace.
Scaffold a Next.js app on Vercel, or a mobile app via Expo with EAS builds — Supabase linked from the first commit.
A board with room to read: description, acceptance criteria and estimate on the card. Tick several, dispatch them in parallel, review them as one queue.
See every active branch and agent at a glance. Open, review, or clean up worktrees without leaving fables.
Script the workspace. Dispatch runs, read the record, and wire fables into CI with a typed SDK and a fables command.
Soft rain, surf, or fire to stay in flow. A quiet companion for long sessions, not another thing shouting for attention.
GitHub, Linear, Jira, Supabase, Vercel, Sentry, PostHog — wired in once, used everywhere across the workspace.
Specs, goals and screenshots live in the workspace and sync to your repo — so every agent works from the same brief.
“You write the wish; the agents write the rest.”
Register your interest and we’ll email you the moment your platform is ready.
Windows & Linux builds coming soon.