Formae's $25 Cloud Plan Is Not Your Infrastructure Bill • The Weighted Average

Small platform teams can test formae Cloud’s September 17 managed-infrastructure launch without adopting another repository-and-pipeline workflow, but its $25 monthly launch price buys management, not the resources being managed. Pair one installation with monthly Claude Pro and the entry subscription stack becomes $45 per month, before the cloud account’s bill, tax, or extra model usage.

The flat price stops at the management boundary

The arithmetic combines two current first-party tariffs. Formae’s pricing page lists $25 per installation per month, while Claude Pro costs $20 when billed monthly. Add them: $25 + $20 = $45. This is one supported assistant-and-management combination, not a mandatory bundle or an all-in deployment estimate. A customer who already pays for the assistant adds only the formae charge; another assistant can have different commercial terms.

That distinction matters because the launch’s strongest promise is operational simplicity. Platform Engineering Labs says the managed service discovers existing infrastructure, records changes, and lets a coding assistant inspect or modify it conversationally. The customer does not have to maintain the infrastructure repository, execution pipeline, or the formae agent itself. The proposed saving is work no longer performed by the team, not free virtual machines hidden inside a subscription.

The product’s architecture explanation keeps the resource boundary explicit. The coding assistant talks over MCP to formae’s service, which routes work to the managed agent. That agent accesses the customer’s AWS, GCP, or Azure account through a granted OIDC role. Infrastructure remains in the customer’s account and on the provider’s bill. A successful conversational deployment can therefore increase that bill even while the management subscription stays flat.

An installation is one managed formae agent instance, with unlimited resources rather than per-seat or per-resource pricing. Unlimited is the billing unit’s property, not an invitation to grant unlimited authority. Before connecting a production account, decide which resources and actions the role should cover. The product says customers can set and change that scope. The sensible pilot uses a disposable environment whose ownership and spending are already clear.

The launch offer also has a clock. The pricing page specifies 7 days of free trial, after which the subscription becomes paid unless cancelled. It says the promotional rate remains for as long as the customer keeps the subscription. Capture those terms before evaluating, including which installation is subscribed. Do not infer that cancellation followed by a later return preserves the offer; the retrieved language only promises continuity while subscribed.

There are meaningful edition differences. The comparison table places enterprise-grade policies and security, non-public hubs, and non-open-source plugins in Enterprise. Cloud lists OIDC, same-day support, and major-cloud and Kubernetes plugins, but not customer plugin installation. A team needing a custom integration should establish its availability before treating the inexpensive Cloud tier as the approved replacement for a broader platform. A low flat rate cannot compensate for a missing operation.

Remove maintenance, not accountability

The Cloud quickstart describes creating a real managed resource through an assistant. That makes the right first experiment concrete: discover a sandbox, make a bounded change, verify the result independently, and examine the recorded history. Then test reversal and a manual change made outside the assistant. These are recommended acceptance checks, not results this publication obtained from operating the service.

The launch claims that formae orders dependencies, checks operations, applies the smallest changes, verifies results, retries, and records what happened. Buyers should ask to see those records during a deliberately interrupted trial. A workflow that recovers cleanly is more valuable than a flawless demonstration that never encountered a partial failure. The evidence should make it possible for the on-call engineer to understand the final state without having participated in the conversation.

This extends our Brig analysis of the difference between an isolation claim and its actual boundary. Here the boundary is a delegated cloud role rather than a local runtime. Avoid translating “no keys to hand over” into “no access risk.” OIDC changes how access is granted; the permissions attached to that grant still determine what the service can do. Restrict them to the work the pilot must prove.

Exit deserves an explicit test too. Formae says cancelling leaves infrastructure in its last live state, with nothing torn down. That protects running resources from subscription-driven deletion, but it does not establish that the customer retains ongoing reconciliation, every historical record, or a ready replacement operating procedure. Ask how to obtain the current definitions and relevant history before leaving. Retaining resources and retaining the ability to manage them are separate acceptance questions.

The strongest objection is that an existing infrastructure workflow already supplies review, history, and recovery at an acceptable maintenance cost. Replacing it can move familiar controls into a supplier’s service without improving delivery. Conversely, a small team spending substantial time maintaining that apparatus has a plausible reason to test the managed alternative. Measure the work actually removed, including the investigation effort retained when an agent misunderstands a request.

There is no retrieved independent productivity measurement that turns the $25 tariff into a guaranteed return. Budget the pilot’s engineering time, assistant usage, real cloud resources, and any Enterprise requirement alongside the subscription. Watch for increased resource creation or lingering test environments; a cheaper management layer is not useful if it makes expensive infrastructure easier to leave running. Use provider-side spending controls rather than expecting the flat subscription to serve that purpose.

Today’s Crusoe lead separates financed infrastructure from delivered service. Formae demands the software version of the same discipline: separate the management offer from what it actually operates. Adopt after a scoped trial demonstrates trustworthy changes, usable evidence, and a workable exit. Delay if custom plugins, policy requirements, or recovery behavior remain unresolved. The attractive entry price earns a test; it does not retire the platform owner’s responsibility.