Why ClusterNest
Boring infrastructure, priced honestly.
Four things decide whether an observability vendor is a good long-term bet: what it runs on, whether it lets you leave, whether the price punishes you for succeeding, and whether the tools your engineers work through can already drive it.
One
Bare metal, not public cloud.
Clusters run on dedicated machines rather than shared cloud instances. That means storage that behaves the same on Tuesday as it did on Monday, no burstable CPU credits to run out of, and no neighbours competing for the same disk, which matters more for search and time series than for almost any other workload.
It also shapes the pricing. A vendor billing per gigabyte on hyperscaler infrastructure is passing on a cloud bill with a margin. Our cost base doesn't scale that way, so our pricing doesn't have to.
Two Standard nodes: 4 vCPU and 32 GiB each, a terabyte of storage across the cluster, Dashboards and coordinator nodes included.
$199.98 / month
Two
We picked the open side, and stayed there.
When Elasticsearch stopped being open source in 2021, we said publicly that it was a loss for the people who'd built on it. In 2022 we launched OpenSearch. In 2024 we retired our Elasticsearch service entirely and migrated the customers still on it.
That's not a values statement, it's a track record with dates on it, and it's the reason we can promise that what you build here isn't hostage to somebody else's licensing decision.
- 2021 Elasticsearch relicensed. We said so, in public.
- 2022 Managed OpenSearch launched.
- 2024 Elasticsearch service retired, customers migrated.
- 2025 Managed Prometheus, Terraform provider, new console.
Three
What you keep if you leave.
The honest test of a managed service is how cheap it is to stop using. Lock-in usually arrives as a format you can't read elsewhere, a query language that only works here, or backups you can't get at. None of those apply.
Open formats
Your data sits in OpenSearch indices and a Prometheus-compatible store. Not a proprietary column format, not a schema only we can read.
Standard query languages
PromQL, OpenSearch DSL, PPL and SQL. Nothing you write here needs rewriting somewhere else, and nothing your team learns is only useful with us.
Backups in your bucket
Snapshots go to S3 storage you own. If we disappeared tomorrow you would still have your data and a standard way to restore it.
An API with everything on it
The console has no privileged operations. Whatever you can click, Terraform and the REST API can do, including leaving.
Four
Your agents already know how to drive it.
Engineers work through coding agents now, and an agent is only as good as the interfaces it has already seen. PromQL, OpenSearch DSL, PPL and SQL have been public and documented for years, so Claude Code, Codex or whichever agent your team has settled on writes working queries against your data on the first attempt, without examples from you. An in-house query language is the opposite: something you re-teach at the start of every session.
The platform is the same story. Every operation lives in a public OpenAPI spec and a Terraform provider, and agents read both formats natively. Give one an API token and it can stand up a cluster, rotate a credential or pull this month's invoice, the same way it would use any other documented API.
The same operations are also served over MCP, for the agents that prefer to be handed tools rather than a specification. Paste the server URL into Claude, Claude Code, Codex or Cursor, log in through the browser once, and the agent works inside your own account and your own organization roles. There is nothing to install and no integration to wait for.
$ export CLUSTERNEST_TOKEN=…
$ claude "give me a Standard OpenSearch 3.7 cluster
called production-logs, reachable only from
203.0.113.0/24"
GET api.clusternest.com/openapi.json 18 endpoints
POST /cluster/opensearch/ 201 created
GET /cluster/opensearch/8/credentials 200 ok
production-logs is provisioning. I'll check back
until it reports running.- Writes PromQL and OpenSearch DSL with no examples from you
- Reads every endpoint from the published OpenAPI spec
- Plans and applies changes through the Terraform provider
- Connects over MCP with a server URL and a browser login
What "managed" means here
Who gets woken up.
"Managed" is used loosely enough to be meaningless, so here it is concretely. You keep cluster admin credentials and full control of configuration. We take the operations.
- We get paged, not you
- Upgrades are ours to plan and run
- Patching happens without a change request from you
- Backups are taken and restore-tested
- Capacity is watched as your indices grow
Ten days, no card.
One Basic cluster, free for ten days. Long enough to point real traffic at it and see whether any of the above holds up.