FAQ

Fair questions, straight answers

What Rollout is, how it compares to the tools you already run, and what happens to your code, your secrets, and your bill.

Product

What is Rollout?
One system with three parts sharing one database: a work tracker (projects, tickets, objectives, custom records), a durable workflow engine (runs, steps, retries, history), and an agent runtime (sandboxed execution on your infrastructure). Agents work your queue through a process you define, and humans keep the gates. The use cases show the three delivery problems teams put it on.
Who is it for?
Small software teams and agencies that want agents doing real delivery work — and the platform engineers who have to answer for where that work runs. If your backlog is bigger than your team and a chat window isn't a process, that's the fit.
How is this different from n8n or Zapier?
Those tools route events between other systems. Rollout runs the work itself: the tracker the agents pull from is built in, the workflow engine is durable (steps, retries, timeouts, full run history), and agent execution is a first-class, sandboxed step — not an HTTP node pointed at someone else's API. There is no sync layer, because the work and the automation live in one database.
How is this different from LangGraph or an agent framework?
Frameworks are libraries — you still build the platform around them: persistence, retries, approvals, multi-tenancy, audit, a UI for the people involved. Rollout is that platform, running. Workflows are visual and durable, runs are recorded, gates are held by humans, and external runtimes can plug in over MCP rather than being rebuilt inside.
We already have Jira and CI. Why move tickets here?
Because the ticket is the agent's context. In Rollout a ticket carries its project, docs, history, and relationships — and delivery is a status machine over those tickets, so moving one to Ready is what dispatches the fleet. Syncing that context out of another tracker is exactly the glue work this replaces. CI keeps doing what CI does; Rollout drives it and records the evidence.
Do I have to write code to use it?
Workflows are built on a visual canvas — triggers, branches, gates, agent steps. When you want code, there's a sandboxed code step (JavaScript on Deno) with explicit resource and network limits. Prompts, models, tools, and iteration caps are node configuration, not a codebase.

Running it

Where do the agents actually run?
In isolated sandboxes on your own Kubernetes or Docker — agent and code steps execute under a hardened sandbox runtime, separated from the rest of your cluster. Managed cloud runners are on the waitlist.
Which AI models can I use?
Bring your own keys. Agent steps run against the provider you configure per node — native provider APIs or OpenRouter for the long tail. Model choice, token budgets, and tool access are part of each node's configuration, so two steps in one workflow can use two different models.
Does my code or data leave my infrastructure?
Deployed on your infrastructure, the platform and the runners are yours — nothing leaves except the model calls you configure, to the providers you chose with your own keys. That boundary is yours to draw, and the run history shows what crossed it.
Can Claude Code or other tools connect?
Yes — Rollout exposes its work graph over MCP, so Claude Code and other MCP clients can read tickets, query records, and drive workflows with the same tenant isolation as the UI. Setup is in the MCP docs.

Security

How are credentials handled?
Secrets live in an envelope-encrypted store and are referenced by name in workflows — values are redacted from logs and run output, and reads fail closed. Connections to external services are scoped per integration.
What does the audit trail cover?
Every run records its steps — inputs, outputs, retries, durations — and every gate records who opened it, for what, and when. Record changes carry actor and timestamp in the activity trail. When someone asks what shipped and why, you open the run. More in the security docs.

Pricing & open source

What does a design partnership include?
The founder builds your first agentic delivery pipeline with you, on your infrastructure. The monthly fee covers the hands-on work and support; where it settles depends on scope. Simple manual invoicing, limited slots.
Can I run Rollout on my own infrastructure today?
Yes — through a design partnership: the founder deploys the platform with you, on your own Kubernetes, and builds your first pipeline. Self-serve self-hosting is planned but not the priority right now. See pricing.
When does Cloud launch, and what will it cost?
Launch pricing is on the pricing page: a per-workspace subscription with unlimited members, an included usage credit, and meters only on what costs money to run — runner minutes, storage, and managed AI tokens (bring your own keys and AI is free on our side). Prepaid, hard-capped, nobody gets surprise-billed. Waitlist workspaces keep the pricing they signed up under.
Will Rollout be open source?
The direction is source-available: self-hosting free for personal use — the whole platform, not a crippled tier — with a commercial license for business use. Paying customers come first, and we won't promise a date we can't hold — an honest "planned" beats a stale "coming soon."

Question not here?

Ask the founder directly, or join the waitlist and hear when the agentic-delivery beta opens.

Prefer to talk it through? Talk to the founder

Prefer to read first? The docs cover every feature on this site.