Cloud Create Project

Create Serverless Project

Create Elastic Cloud Serverless projects using the Serverless REST API. Use the cloud-manage-project skill for day-2 operations like listing, updating, or deleting projects.

Prerequisites and permissions

• Ensure EC_API_KEY is configured. If not, run cloud-setup skill first.
• Creating projects requires a Cloud API key with Admin or Organization owner role.
• This skill does not perform a separate role pre-check. Attempt the requested operation and let the API enforce

authorization. If the API returns an authorization error (for example, 403 Forbidden), stop and ask the user to verify the provided API key permissions.

Manual setup fallback (when cloud-setup is unavailable)

If this skill is installed standalone and cloud-setup is not available, instruct the user to configure Cloud environment variables manually before running commands. Never ask the user to paste API keys in chat.

Note: If EC_API_KEY is missing, or the user does not have a Cloud API key yet, first check whether the user has

an Elastic Cloud account. If not, propose starting a free trial at

Elastic Cloud free trial — 14 days of full access with no credit card

required. Once registered, direct the user to generate a key at

Elastic Cloud API keys, then configure it locally using the steps below.

Preferred method (agent-friendly): create a .env file in the project root:

EC_API_KEY=your-api-key
EC_BASE_URL=https://api.elastic-cloud.com

All cloud/* scripts auto-load .env from the working directory.

Alternative: export directly in the terminal:

export EC_API_KEY="<your-cloud-api-key>"
export EC_BASE_URL="https://api.elastic-cloud.com"

Terminal exports may not be visible to sandboxed agents running in separate shell sessions, so prefer .env when using an agent.

Critical principles

• Never display secrets in chat. Do not echo, log, or repeat API keys, passwords, or credentials in conversation

messages or agent thinking. Direct the user to the .elastic-credentials file instead. The admin password must never appear in chat history, thinking traces, or agent output.

• Confirm before creating. Always present the project configuration to the user and ask for confirmation before

running the creation script.

• Admin credentials are for API key creation only. The script saves the admin password to .elastic-credentials

for bootstrapping a scoped API key. The admin user has full privileges and cannot be modified in serverless. Never use admin credentials for direct Elasticsearch operations (querying, indexing, etc.) — always create a scoped API key first (see Step 8). The load-credentials command excludes admin credentials by default — use --include-admin only during Step 7/8, then reload without it once the API key is created. Never read or display the contents of .elastic-credentials in chat.

• Recover lost credentials. If the script fails to write .elastic-credentials (disk full, permissions, etc.), the

save may be incomplete. Check .elastic-credentials for the password first. If missing, use the cloud-manage-project skill's reset-credentials command to generate a new password.

• Region is permanent. A project's region cannot be changed after creation.
• Prefer automatic readiness checks. Pass --wait to the creation script so it polls until the phase changes from

initializing to initialized. Only fall back to manually polling the status endpoint if --wait is unavailable.

Project types

Project type inference

Map the user's request to the correct --type value:

Do not silently default to any type. If the user does not specify a type, infer it from the conversation context (for example, discussing log ingestion suggests observability, discussing detections or SIEM suggests security, discussing search or vector workloads suggests elasticsearch). Always present the inferred type to the user and ask for confirmation before proceeding. If context is insufficient to infer a type, ask the user to choose.

Product tiers

Observability and security projects support a --product-tier flag. Default to complete unless the user explicitly requests a different tier.

Elasticsearch projects do not have a product tier — use --optimized-for instead.

Sensible defaults

Present these defaults to the user before creation. Ask if they want to use or change them:

Project type must be confirmed with the user — do not assume a default. See "Project type inference" above.

Always use --optimized-for general_purpose unless the user explicitly requests vector. Do not proactively offer the vector option.

If the user does not specify a name, ask for one — it is required.

Workflow: Create a project

Project Creation:
- [ ] Step 1: Verify API key is set
- [ ] Step 2: Present defaults and confirm with user
- [ ] Step 3: List available regions (optional)
- [ ] Step 4: Create the project
- [ ] Step 5: Save credentials and endpoints
- [ ] Step 6: Wait for project to initialize
- [ ] Step 7: Set environment variables
- [ ] Step 8: Recommend creating a scoped API key

Step 1: Verify API key is set

echo "${EC_API_KEY:?Not set}"

If EC_API_KEY is not set, run the cloud-setup skill first to configure authentication and defaults.

Step 2: Present summary and confirm with user

Before presenting the summary, ensure the project type has been explicitly confirmed by the user. If no type was specified, infer one from the conversation context and propose it. If the context is ambiguous, ask the user to choose from elasticsearch, observability, or security.

Always show a confirmation summary before creating. Include different fields depending on project type:

Elasticsearch project:

Project Summary:
  Type:          elasticsearch
  Name:          my-project
  Region:        gcp-us-central1

Observability project:

Project Summary:
  Type:          observability
  Name:          my-project
  Region:        gcp-us-central1
  Product tier:  complete

Security project:

Project Summary:
  Type:          security
  Name:          my-project
  Region:        gcp-us-central1
  Product tier:  complete

Ask the user to confirm or override any values before proceeding.

Step 3: List available regions (optional)

python3 skills/cloud/create-project/scripts/create-project.py list-regions

The output is grouped by cloud provider (AWS, Azure, GCP) and sorted alphabetically. Regions marked with * do not support project creation.

Step 4: Create the project

python3 skills/cloud/create-project/scripts/create-project.py create \
  --type elasticsearch \
  --name "my-project" \
  --region gcp-us-central1 \
  --optimized-for general_purpose \
  --wait

Always pass --optimized-for general_purpose for Elasticsearch projects. Only use vector if the user explicitly requests it.

For observability and security projects, pass --product-tier complete unless the user explicitly requests a different tier.

Always pass --wait so the script automatically polls until the project is ready.

Step 5: Save credentials and endpoints

The script automatically writes credentials to .elastic-credentials in the working directory. The password is redacted from the JSON output on stdout.

If saving succeeds, tell the user:

Credentials saved to .elastic-credentials — open that file to retrieve your password.

Do not read, cat, or display the contents of .elastic-credentials in chat.

If saving fails, the script prints an error to stderr. Check whether .elastic-credentials exists and contains a password (a partial write is possible). If the password is missing or the file does not exist, immediately run the cloud-manage-project skill's reset-credentials command to generate a new password.

The creation response also contains:

• Project ID — needed for all subsequent operations
• Cloud ID — for client libraries
• Elasticsearch and Kibana endpoints — safe to display in chat

The admin credentials are for initial bootstrap only. Recommend creating a scoped API key for ongoing access (Step 8).

Step 6: Wait for project to initialize

When --wait is passed (recommended), the script polls automatically until the project phase becomes initialized. No manual polling is needed.

If the agent ran without --wait, poll manually:

python3 skills/cloud/create-project/scripts/create-project.py status \
  --type elasticsearch \
  --id <project-id>

Repeat until phase changes from initializing to initialized.

Step 7: Set environment variables

The creation script saves credentials and endpoints to .elastic-credentials with the project name in the header. Load them into the current shell with `--include-admin` so admin credentials are available for API key creation in Step 8:

eval $(python3 skills/cloud/manage-project/scripts/manage-project.py load-credentials \
  --name "<project-name>" --include-admin)

This sets ELASTICSEARCH_URL, KIBANA_URL, any project-type specific endpoints (APM_URL, INGEST_URL), and the admin ELASTICSEARCH_USERNAME/ELASTICSEARCH_PASSWORD needed to bootstrap an API key.

Step 8: Create a scoped API key

The admin user has full privileges and cannot be modified in serverless projects. Do not proceed with Elasticsearch operations using admin credentials. Create a scoped Elasticsearch API key with only the permissions the user needs.

If the elasticsearch-authn skill is available, use it for API key creation — it covers the full lifecycle (create, grant, invalidate, query) and handles scoping privileges correctly. If the skill is not installed, ask the user to either install it or create the API key manually through Kibana > Stack Management > API keys. After creation, save the API key to .elastic-credentials using the project-specific header format (see manage-project skill's "Credential file format" section), then reload without `--include-admin` to drop admin credentials from the environment:

eval $(python3 skills/cloud/manage-project/scripts/manage-project.py load-credentials \
  --name "<project-name>")

Examples

Create an Elasticsearch project with defaults

python3 skills/cloud/create-project/scripts/create-project.py create \
  --type elasticsearch \
  --name "my-search-project" \
  --region gcp-us-central1 \
  --optimized-for general_purpose \
  --wait

Create an observability project

python3 skills/cloud/create-project/scripts/create-project.py create \
  --type observability \
  --name "prod-o11y" \
  --region aws-eu-west-1 \
  --product-tier complete \
  --wait

Create a security project

python3 skills/cloud/create-project/scripts/create-project.py create \
  --type security \
  --name "siem-prod" \
  --region gcp-us-central1 \
  --product-tier complete \
  --wait

Guidelines

• Run the cloud-setup skill first if EC_API_KEY is not set.
• Always confirm the project configuration with the user before creating.
• Never display passwords or API keys in chat. Direct the user to .elastic-credentials.
• Never silently default to a project type. Infer from context and confirm with the user.
• Default to general_purpose optimization. Only use vector if the user explicitly requests it.
• Default to complete product tier for observability and security projects. Only use logs_essentials or essentials

if the user explicitly requests it.

• Always pass --wait so the script polls until the project is ready.
• If credential saving fails, immediately reset credentials using the cloud-manage-project skill.
• After creation, recommend creating a scoped API key instead of relying on admin credentials.
• Region cannot be changed after creation — confirm the choice before proceeding.

Script reference

Environment variables

Additional resources

• For full API details, request/response schemas, and project-type options, see

references/api-reference.md

Elastic Cloud Serverless API Reference

Reference for the Elastic Cloud Serverless project management API. Base URL: https://api.elastic-cloud.com.

All requests require an API key in the Authorization header:

Authorization: ApiKey <your-api-key>

Endpoints

Projects

Where {type} is one of: elasticsearch, observability, security.

Regions

Traffic Filters

Create project request bodies

Elasticsearch

{
  "name": "my-es-project",
  "region_id": "gcp-us-central1",
  "optimized_for": "general_purpose",
  "search_lake": {
    "search_power": 100,
    "boost_window": 30
  }
}

optimized_for options:

• general_purpose — suitable for full-text search, sparse vectors, compressed dense vectors. Default when creating

from the UI.

• vector — for uncompressed dense vectors with high dimensionality. Impacts VCU consumption.

Observability

{
  "name": "my-obs-project",
  "region_id": "gcp-us-central1",
  "product_tier": "complete"
}

Security

{
  "name": "my-sec-project",
  "region_id": "gcp-us-central1",
  "product_types": [
    { "product_line": "security", "product_tier": "complete" },
    { "product_line": "cloud", "product_tier": "complete" },
    { "product_line": "endpoint", "product_tier": "complete" }
  ]
}

Product lines: security, cloud, endpoint. Product tiers: complete, essentials.

Creation response

All project types return the same response structure:

{
  "id": "abc123def456abc123def456abc12345",
  "name": "my-project",
  "alias": "my-project-bdc0f7",
  "region_id": "gcp-us-central1",
  "cloud_id": "bXktcHJvamVjdDo...",
  "type": "elasticsearch",
  "metadata": {
    "created_at": "2025-01-15T10:30:00Z",
    "created_by": "1014289666002276",
    "organization_id": "198583657190"
  },
  "endpoints": {
    "elasticsearch": "https://my-project-bdc0f7.es.us-east-1.aws.elastic.cloud",
    "kibana": "https://my-project-bdc0f7.kb.us-east-1.aws.elastic.cloud"
  },
  "credentials": {
    "username": "admin",
    "password": "REDACTED"
  }
}

Observability projects also include apm and ingest endpoints. Security projects include an ingest endpoint.

The credentials block is only returned on creation and reset. Save the password immediately.

Project status

{
  "phase": "initializing"
}

Phases:

Regions

Common regions for serverless projects:

AWS

GCP

Azure

Use GET /api/v1/serverless/regions for the complete, up-to-date list. Only regions with project_creation_enabled: true accept new projects.

Error responses

All errors follow this format:

{
  "errors": [
    {
      "message": "Explanatory error message",
      "code": "code.error"
    }
  ]
}

Traffic filters

Traffic filters restrict network access to projects. Attach them during creation or using PATCH:

{
  "traffic_filters": [{ "id": "traffic-filter-id" }]
}

Filter types: ip (IP/CIDR rules) or vpce (VPC endpoint IDs).