Roadmap
Where Moa is headed. These are planned directions, not shipped features — the goal is to let you run Moa on your own terms, on any model, and reach it from wherever you already work.
Run on your own dedicated machine#
A fully isolated, dedicated Moa deployment — your own instance on your own compute, rather than a workspace on shared infrastructure. You'd control where it runs and where its data lives, with the same engine and dashboard on top.
- Dedicated compute per team, isolated end to end.
- Deploy into your own cloud or hardware.
- Your data and secrets never leave your environment.
Bring your own model keys — Claude and OpenAI#
Today the agent runs on Claude. Next is multi-provider: connect your own Claude and OpenAI keys, and choose which model powers which workflow — a cheaper model for triage, a stronger one for solving. You bring the keys; usage bills to you.
Drive Moa from Slack#
A Slack integration so you can trigger and follow runs without leaving the conversation. Mention Moa with a task in a thread, and it streams the live step timeline back into that thread — the same deterministic engine, a chat-native surface.
Embed Moa anywhere#
Moa is already API-first — the dashboard, CLI, and webhooks are all clients of one server. The plan is to make that easy to embed: drop Moa's controls and run views into your own internal dashboard or tools via a clean SDK, so it lives where your team already works.
Delegate heavy tasks from anywhere#
Hand off long-running or heavy work — a big refactor, a repo-wide audit, a migration — to Moa from any surface (your editor, CI, Slack, your own app) and let it run in the background. You delegate the task and get back the result, instead of babysitting a terminal.
And more#
- Streamed cross-workflow activity — one live view of everything Moa is doing.
- Richer memory capture and file-hotspot intelligence.
- Issue dependency gating — "blocked by" awareness before a run starts.