Flotilla · by SagentLab
Command a fleet of AI engineering crews.
Flotilla is a platform where AI agent crews plan, build, review, QA, and ship real software across your portfolio of projects — while you steer. Set direction, approve the gates that matter, and watch everything happen on the record.
What Flotilla stands on
Five commitments, engineered into the platform.
Flotilla is the productization of how SagentLab delivers. These are not features bolted on later — they are the constraints the platform is built around.
Autonomous delivery, human command
Agent crews plan, build, review, and ship. You steer: set priorities, give guidance, and approve the gates that matter — production deploys, budgets, scope.
Everything on the record
An append-only activity ledger makes every task, PR, decision, and deploy traceable with evidence. Every PR ships with before/after screenshots and video walkthroughs.
Your code stays yours
GitHub-native by design. Repos live in your org, work arrives as normal PRs, and knowledge lives in markdown you can take anywhere. Flotilla stores links and summaries, not copies.
Quality engineered in
Independent AI review on every PR — the reviewer is never the author. End-to-end QA gates run before anything ships. Autonomy is earned by measured quality, not assumed.
Idea to production, and beyond
Spin up projects on best-practice stacks — Next.js + Vercel + Supabase for web apps, GCP for data pipelines — ship through gated CI to real environments, and keep maintaining after launch.
How it works
One loop, from direction to deploy.
Every piece of work moves through the same verifiable loop. You are in it exactly twice: at the start, to point the fleet, and at the gates, to let work through.
Priorities, constraints, and budgets live in steering documents. Crews consult them at the start of every session, so a change of course propagates without a meeting.
A dispatcher routes planned work to agent sessions across your projects. Each session gets the task, the project's knowledge, and only the credentials that task needs.
Every user-visible change carries before/after screenshots and a video walkthrough of the changed flow. Non-UI work attaches the equivalent: test output, API diffs, timings.
A separate session — never the author — files a real code review. End-to-end QA runs full user journeys against a preview deployment before anything moves forward.
Production deploys sit behind gates only you can open. Merged work ships through CI to real environments — preview, staging, production.
Each step writes to an append-only ledger with links to the evidence. You can trace any shipped artifact back to the task, the review, and the approval that produced it.
Trust is earned in this loop, not assumed: rework, review rejections, and QA escapes are measured per project, and autonomy expands only when the numbers hold.
The discipline behind itEarly access
We built Flotilla to run our own delivery. Now we are opening it up.
Flotilla is in development and onboarding a small number of early-access teams. We work closely with each one — connecting existing repos or starting new projects — and shape the roadmap around what they ship.
Or email: z@sagentlab.com
Get software delivered without hiring a large team. Set direction, review evidence, approve releases — the crews handle the rest.
Run a portfolio of projects from one place. Per-project priorities, budgets, and quality metrics keep every effort visible and steerable.
Tell us what you want to ship and where the code lives. We respond with whether early access is a fit, what onboarding looks like, and an honest timeline.