Connect Elastic Cloud to Claude: Administer Users and API Access
A complete engineering guide to securely connecting Elastic Cloud to Claude using an MCP server. Automate deployment scaling, RBAC, and user provisioning.
If your team needs to connect Elastic Cloud to Claude to administer user access, scale Elasticsearch tiers, or automate API key provisioning, you need a Model Context Protocol (MCP) server. This server acts as the translation layer between Claude's natural language tool calls and Elastic Cloud's REST APIs. You can either build and maintain this infrastructure yourself, or use a managed integration platform like Truto to dynamically generate a secure, authenticated MCP server URL. If your team uses ChatGPT, check out our guide on connecting Elastic Cloud to ChatGPT or explore our broader architectural overview on connecting Elastic Cloud to AI Agents.
Giving a Large Language Model (LLM) read and write access to a sprawling infrastructure platform like Elastic Cloud is a significant engineering challenge. You must handle complex authentication lifecycles, map heavily nested deployment schemas to MCP tool definitions, and deal with strict state-management rules. Every time the upstream API changes its cluster sizing logic or deprecates a template, you have to update your server code, redeploy, and test the integration.
This guide breaks down exactly how to use Truto to generate a secure, managed MCP server for Elastic Cloud, connect it natively to Claude Desktop, and execute complex infrastructure workflows using natural language.
The Engineering Reality of the Elastic Cloud API
A custom MCP server is a self-hosted integration layer. While the open MCP standard provides a predictable way for models to discover tools, the reality of implementing it against Elastic Cloud's API is painful. You are not just integrating a simple CRUD application - you are interfacing with a complex orchestration engine.
If you decide to build a custom MCP server for Elastic Cloud, here are the specific integration challenges you will face:
Heavily Nested Topology and Tier Sizing
Updating an Elastic Cloud deployment is not a matter of sending a flat JSON patch payload. If you want to upsize a cluster, the API requires you to pass the target sizes within a deeply nested resources block that specifies the topology arrays for hot_content, warm, cold, frozen, master, and ml tiers. An LLM has no inherent context on which tiers are enabled or what the valid memory-to-zone ratios are. Without strict JSON Schema validation injected into the MCP tool definition, the model will hallucinate flat update payloads that the Elastic API will outright reject.
Asynchronous State and Pending Plans
Elastic Cloud operations are highly asynchronous. When you restart an Elasticsearch resource or apply a topology change, the API returns immediately with an accepted response, but the actual infrastructure change enters a pending plan state. If an LLM assumes 200 OK means the cluster is fully restarted, subsequent tool calls expecting the new state will fail. Your MCP layer must account for these asynchronous plan executions.
Complex RBAC and Role Assignments
Provisioning a user or generating an API key in Elastic Cloud requires precise role assignments scoped across platform, organization, deployment, and project dimensions. The API enforces strict validation on these role combinations. You cannot simply tell the API to "make this user an admin." You have to provide the exact role mapping array. An LLM needs explicit schemas detailing these enumerated role types to prevent 400 Bad Request errors during provisioning.
Generating the Elastic Cloud MCP Server
Truto abstracts away the schema mapping and token management by dynamically generating an MCP server from the Elastic Cloud integration's raw API documentation. You can spin up this server through the UI or programmatically via the API.
Method 1: Via the Truto UI
For IT admins or DevOps engineers who want to quickly provide an MCP connection to their team without writing a script:
- Navigate to the Integrated Accounts page in your Truto dashboard.
- Select your active Elastic Cloud connection.
- Click the MCP Servers tab.
- Click Create MCP Server.
- Configure the server settings (assign a descriptive name, filter allowed methods to
readorwrite, and set an optional expiration date). - Click Generate and copy the resulting MCP server URL.
Method 2: Via the Truto API
For platform engineering teams automating infrastructure access, you can generate MCP servers programmatically. This validates that tools are available, stores the hashed token in a distributed key-value store for edge validation, and returns a ready-to-use URL.
curl -X POST https://api.truto.one/integrated-account/{integrated_account_id}/mcp \
-H "Authorization: Bearer YOUR_TRUTO_API_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"name": "Elastic Cloud Ops - Claude Access",
"config": {
"methods": ["read", "write"],
"require_api_token_auth": false
},
"expires_at": "2025-12-31T23:59:59Z"
}'The API returns the fully configured MCP server URL:
{
"id": "mcp_2a3b4c5d",
"name": "Elastic Cloud Ops - Claude Access",
"config": { "methods": ["read", "write"] },
"expires_at": "2025-12-31T23:59:59.000Z",
"url": "https://api.truto.one/mcp/e7f8g9h0..."
}Connecting the MCP Server to Claude
Once you have your Truto MCP URL, connecting it to Claude requires no additional middleware. You can configure this directly in the Claude interface or via a local configuration file for Claude Desktop.
Method 1: Via the Claude UI
If you are using the Claude web interface or enterprise workspace settings:
- Open Claude and navigate to Settings -> Integrations.
- Locate the Add MCP Server or Custom Connector option.
- Paste the Truto MCP URL (e.g.,
https://api.truto.one/mcp/e7f8g9h0...). - Click Add. Claude will immediately perform a handshake, run the
tools/listprotocol, and make the Elastic Cloud tools available in the context window.
Method 2: Via Manual Configuration File
For developers using the Claude Desktop application locally, you can map the URL directly in your claude_desktop_config.json file using the official remote SSE transport.
Open your configuration file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Add the Truto MCP server using the @modelcontextprotocol/server-sse package:
{
"mcpServers": {
"elastic-cloud-ops": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-sse",
"https://api.truto.one/mcp/e7f8g9h0..."
]
}
}
}Restart Claude Desktop. The application will initialize the server and load the Elastic Cloud operations into your agent's available toolset.
Hero Tools for Elastic Cloud Administration
Truto automatically translates Elastic Cloud's API endpoints into strongly typed MCP tools. The query and body schemas instruct Claude exactly how to structure the deeply nested JSON payloads required by Elastic Cloud. Here are 6 high-leverage tools available for your agent.
list_all_elastic_cloud_deployments
Fetches an inventory of all deployments across your Elastic Cloud environment. This is the foundational discovery tool Claude uses to map deployment IDs to human-readable cluster names.
"Claude, list all the current Elastic Cloud deployments in our environment and give me a summary table of their names, IDs, and current versions."
elastic_cloud_deployment_elasticsearch_tiers_patch
Updates the size and/or zone count of one or more Elasticsearch topology elements (tiers) within a Deployment. Claude relies on the strict schema provided by Truto to pass the correct memory increments for hot_content, warm, or cold tiers.
"Claude, upsize the hot tier of the production-search deployment to 8GB per zone and increase the zone count to 3 for high availability."
elastic_cloud_deployment_resources_restart_stateless
Restarts a stateless resource like Kibana or Enterprise Search. If the resource is active, this applies a cluster reboot; if inactive, it starts the resource with the most recent successful plan.
"The Kibana instance for our logging deployment is unresponsive. Please issue a stateless restart command for the Kibana resource on deployment ID d3a4..."
create_a_elastic_cloud_api_key
Provisions a new API key. Because API keys are only returned once in the response, Claude can generate the key and immediately securely output it to the user or pass it to another tool.
"Create a new Elastic Cloud API key named 'Datadog-Metrics-Ingest'. Give it an expiration of 90 days and output the resulting key value."
create_a_elastic_cloud_organization_invitation
Creates or refreshes organization invitations. This tool allows Claude to automate user onboarding by dispatching invites to specific email addresses and assigning initial roles.
"Invite john.doe@example.com to our Elastic Cloud organization and assign them the standard developer role."
list_all_elastic_cloud_deployment_elasticsearch_resources
Retrieves detailed status, topology, and configuration info about an Elasticsearch resource belonging to a given Deployment. Critical for diagnosing cluster health or checking configuration drift.
"Pull the configuration details for the Elasticsearch resource on the staging deployment. What is the current configuration of its keystore and remote clusters?"
To view the complete inventory of available Elastic Cloud operations, schemas, and required parameters, visit the Elastic Cloud integration page.
Real-World Workflows in Action
When Claude is equipped with an Elastic Cloud MCP server, it can orchestrate complex, multi-step infrastructure tasks that would normally require digging through the Elastic Cloud console or writing custom Python scripts. Here is how specific personas use these workflows in practice.
Workflow 1: Upsizing an Elasticsearch Cluster
Persona: Site Reliability Engineer (SRE)
During a major traffic spike, an SRE needs to scale up the primary logging cluster rapidly without navigating a complex UI.
"Claude, find the deployment named 'Prod-Log-Ingest' and scale its hot tier to 16GB of RAM across 3 zones."
- Claude calls
list_all_elastic_cloud_deploymentsto search for the deployment named 'Prod-Log-Ingest' and extracts itsdeployment_idand theref_idfor the Elasticsearch resource. - Claude evaluates the current tier sizes by calling
list_all_elastic_cloud_deployment_elasticsearch_tiersusing the retrieved IDs. - Claude calls
elastic_cloud_deployment_elasticsearch_tiers_patch, structuring the nested JSON body to target thehot_contenttier with the new size and zone count. - Claude returns a summary to the SRE confirming that the topology update plan has been submitted successfully.
sequenceDiagram
participant SRE as "SRE"
participant Claude as "Claude"
participant MCP as "Truto MCP Server"
participant Elastic as "Elastic Cloud API"
SRE->>Claude: Scale 'Prod-Log-Ingest' hot tier to 16GB / 3 zones
Claude->>MCP: Call list_all_elastic_cloud_deployments
MCP->>Elastic: GET /deployments
Elastic-->>MCP: Returns list (ID: d92f...)
MCP-->>Claude: JSON result
Claude->>MCP: Call elastic_cloud_deployment_elasticsearch_tiers_patch(id, hot_content config)
MCP->>Elastic: PATCH /deployments/{id}/elasticsearch/{ref_id}/tiers
Elastic-->>MCP: 200 OK (Plan Pending)
MCP-->>Claude: JSON result
Claude-->>SRE: Confirms cluster scaling is initiatingWorkflow 2: Automating API Key Provisioning for a Vendor
Persona: Security / Identity Admin
A security admin needs to issue a temporary API key for a compliance auditor, but doesn't want to leave long-lived credentials sitting in the system.
"Generate an API key named 'Q3-Compliance-Audit' for the security team. Make sure it expires in exactly 14 days, then show me the key."
- Claude calculates the ISO 8601 timestamp for 14 days from the current date.
- Claude calls
create_a_elastic_cloud_api_key, passing the description "Q3-Compliance-Audit" and the calculated expiration date. - Elastic Cloud generates the key and returns it in the payload.
- Claude presents the one-time key string to the administrator.
Workflow 3: Orchestrating a Stateless Resource Restart
Persona: DevOps Engineer
A DevOps engineer is alerted that Kibana is throwing 502 errors and needs to reboot the stateless component without taking down the underlying Elasticsearch cluster.
"Kibana is hanging on the 'Data-Science-Dev' deployment. Please restart the stateless Kibana resource."
- Claude calls
list_all_elastic_cloud_deploymentsto resolve the string "Data-Science-Dev" to a specificdeployment_id. - Claude calls
list_all_elastic_cloud_deployment_kibana_resourcesto identify theref_idof the specific Kibana instance attached to that deployment. - Claude calls
elastic_cloud_deployment_resources_restart_stateless, specifying theresource_kindaskibanaalong with the IDs. - Claude informs the engineer that the cluster reboot command was successfully issued.
Security and Access Control
Handing infrastructure control over to an AI agent requires strict guardrails. Truto's MCP implementation provides four critical layers of security configuration that you can define when generating the server:
- Method Filtering: Use the
methodsarray to restrict the MCP server entirely to["read"]operations (likelistandget). This guarantees that Claude can audit deployments but cannot mutate infrastructure or delete users. - Tag Filtering: Use the
tagsarray to scope tools down to specific functional areas. For example, filtering by["users"]ensures the agent only has access to invitation and role mapping tools, preventing it from touching deployment topologies. - Secondary Authentication (
require_api_token_auth): By default, the MCP token embedded in the URL grants access. By settingrequire_api_token_auth: true, the calling client must also pass a valid Truto API token in the Authorization header. This means possessing the URL is not enough to execute tools. - Automated Expiration (
expires_at): Set an absolute ISO datetime for the server's death. Once expired, the underlying key-value edge entries and database records are automatically scrubbed, instantly revoking Claude's access to the environment.
Handling Rate Limits and Retries
When automating bulk infrastructure checks (like scraping the resource settings across 50 different Elastic Cloud deployments), Claude may hit Elastic Cloud's API rate limits.
It is critical to understand that Truto acts as a strict pass-through layer for rate limiting. Truto does not retry, throttle, or apply backoff algorithms on rate limit errors.
When the Elastic Cloud API rejects a request with an HTTP 429 Too Many Requests status, Truto passes that 429 directly back to the calling client (Claude). Truto actively normalizes the upstream rate limit information into standardized IETF headers (ratelimit-limit, ratelimit-remaining, ratelimit-reset).
The caller (the MCP client, Claude Desktop, or your custom agent framework) is entirely responsible for inspecting these headers and implementing the appropriate backoff-and-retry logic. Do not expect the MCP server to absorb these errors for you.
Start Automating Your Infrastructure
Connecting Elastic Cloud to Claude via a managed MCP server removes the friction of building OAuth flows, maintaining OpenAPI schema mappings, and writing custom wrapper APIs. By leveraging documentation-driven tool generation, you ensure your AI agents always have access to the most accurate, strongly typed API endpoints available.
Whether you are restricting access to read-only audits or allowing full infrastructure scaling workflows, the MCP architecture gives your team a secure, auditable way to interact with Elastic Cloud using natural language.
FAQ
- Can I restrict Claude to only read data from Elastic Cloud?
- Yes. When configuring the MCP server in Truto, you can set the `methods` configuration to `["read"]`. This limits the exposed tools to `get` and `list` operations, ensuring the agent cannot mutate or delete infrastructure.
- Does Truto automatically retry failed requests if Elastic Cloud rate-limits the agent?
- No. Truto does not retry, throttle, or apply backoff logic. When Elastic Cloud returns an HTTP 429, Truto passes the error back to the caller alongside standardized IETF headers (ratelimit-limit, ratelimit-remaining, ratelimit-reset). The calling agent is responsible for retries.
- How do I revoke an MCP server's access to my Elastic Cloud account?
- You can delete the MCP server directly from the Truto UI or via the API. Alternatively, you can configure an `expires_at` timestamp during creation, which will automatically scrub the token from the edge network and database when the time is reached.