AI readiness
Engineers work through coding agents now, and an agent is only as good as the interfaces it has already seen. Everything here is one it has: open query languages, a published spec, an MCP connector on the platform, and an MCP server on the clusters themselves.
Agent compatibility is mostly a property of open standards rather than a feature anyone builds. Two of these three layers were true before MCP existed, which is why there is nothing to install and no integration to wait for.
PromQL, OpenSearch DSL, PPL and SQL have been public and documented for years, so an agent writes working queries against your data on the first attempt. An in-house query language is the opposite: something you re-teach at the start of every session.
Every operation is in a published OpenAPI spec and a Terraform provider. Both are formats an agent reads unaided, which is why this worked before either MCP integration existed.
An MCP connector for the platform and an MCP server inside each OpenSearch cluster. The connector manages clusters, members and invoices, and runs each cluster’s tools for you. The server searches your indices and reads their mappings.
The ClusterNest MCP connector exposes the REST API’s operations as tools. Paste the server URL and client ID into Claude, Claude Code, Codex or Cursor, log in through the browser once, and the client refreshes its own token from then on. There is no client secret and nothing to pre-register.
| Server URL | https://api.clusternest.com/mcp |
|---|---|
| OAuth client ID | mcp |
| Client secret | None. It is a public client. |
| Transport | Streamable HTTP |
| Cost | Included on every tier |
The cluster, as tools
OpenSearch ships an MCP server of its own inside the ML Commons plugin, from version 3.3.0. Ask your agent to turn it on, and the ClusterNest MCP connector registers three read-only tools: list the indices, read their mappings, and search them. The agent runs them through the same connector, with your role, so there is no second server to add and no password to hand over.
It is off until you turn it on, and the registered tools are the whole of what it offers. A client that can add servers of its own can also connect to it directly, with a short-lived token from the connector.
$ clusternest enable_opensearch_cluster_mcp
production-logs enabled
$ clusternest list_opensearch_cluster_mcp_tools
ListIndexTool read-only
IndexMappingTool read-only
SearchIndexTool read-onlyOne connector does both. The agent stands the cluster up, turns on its MCP server and searches it, so the work that used to be a console tab, a terminal and a Dashboards window is one conversation.
The connector receives the tool calls your agent makes and what the cluster answers. The cluster answers from the indices in it. Neither receives the conversation that produced the call.
$ claude "stand up a Standard cluster for
the checkout logs and turn on search"
clusternest create_opensearch_cluster
checkout-logs · standard creating
checkout-logs available
clusternest enable_opensearch_cluster_mcp
checkout-logs enabled
checkout-logs is up and searchable. Point
your shipper at it, and ask me about the
logs once they arrive.With one connector that reaches the data, anyone in the organization can ask their agent about it from any MCP client, the Claude app on their phone included, with the access their role already gives them.
“What started returning 500s on checkout in the last hour?” asked from the Claude app on your phone. No laptop, no VPN, no Dashboards login.
Members sign in with their ClusterNest account. A team with no identity provider, or one its agents can’t log in through, still gets an agent that can search its logs.
Support and product members invited as readers ask in plain language and get answers from the same data.
Invite them as readers and they query through their own assistant, with no password to hand over. Remove them and they can no longer sign in.
Ask for a cluster, load an export into it, ask for the top error patterns, and have the agent delete it afterwards. The whole lifecycle is one conversation.
Runbooks and postmortems for a team, or documents, receipts and notes for one person. Whatever you index, your assistant can look up from any device.
Giving an agent access to production infrastructure is a real decision, so the limits are worth stating plainly. They are the ones already enforced for people, applied to the same accounts.
The full account model, including what each organization role can do, is on security and compliance.
Ten days, one Basic cluster, no card. Long enough to connect a client, provision a cluster from a prompt and decide whether working this way suits your team.