An AI-powered assistant that guides you from initial requirements to a running OpenSearch search setup. It collects a sample document, gathers preferences, plans a search architecture, and executes the plan — creating indices, ML models, ingest pipelines, and a local search UI — with optional deployment to Amazon OpenSearch Service or Serverless.
Works with Claude Code, Cursor, Kiro, and any agent that supports the Agent Skills specification.
| Skill | Category | Description |
|---|---|---|
| opensearch-launchpad | General | Get started with OpenSearch. Guides you through semantic, hybrid, neural, and agentic search setup with local execution and optional AWS deployment. |
More skills coming soon — contributions welcome! See Contributing a New Skill.
Install any skill into your project using npx skills:
npx skills add opensearch-project/opensearch-launchpadThis discovers skills under skills/ and symlinks them into your agent's skill directory (.claude/skills/, .cursor/skills/, etc.). Works with Claude Code, Cursor, OpenCode, Codex, and many more.
# Install to a specific agent
npx skills add opensearch-project/opensearch-launchpad -a claude-code
# Install globally (available across all projects)
npx skills add opensearch-project/opensearch-launchpad -g
# Install to all detected agents
npx skills add opensearch-project/opensearch-launchpad --all
# List available skills before installing
npx skills add opensearch-project/opensearch-launchpad --listAfter installing, try:
"I want to build a semantic search app with 10M docs"
Your agent reads the skill instructions and runs the scripts directly — no MCP server required.
OpenSearch Launchpad Power — Add this power source URL in Kiro to get started.
- Open Kiro
- Go to Powers panel
- Click Add Power and paste:
https://github.com/opensearch-project/opensearch-launchpad/tree/main/kiro/opensearch-launchpad - Kiro reads
POWER.mdand connects the MCP server automatically — no local clone required.
| Path | IDEs | How it connects |
|---|---|---|
| Agent Skill | Claude Code, Cursor, Kiro, OpenCode, Codex | Agent reads SKILL.md and runs scripts directly via the terminal |
| Kiro Power | Kiro | Kiro runs the MCP server (opensearch-launchpad) which exposes phase tools |
The Agent Skill path uses standalone scripts with zero dependency on the MCP server or opensearch_orchestrator package. The Kiro Power path is maintained for backward compatibility with existing Kiro Power installations.
- Python 3.11+ and
uvinstalled - Docker installed and running (Download Docker)
- For AWS deployment (optional): AWS credentials configured — see AWS Setup
OpenSearch Launchpad walks you through five phases to build a production-ready search solution:
| Phase | What happens |
|---|---|
| 1. Sample Document | Provide a sample document (built-in IMDB dataset, local file, URL, existing index, or paste JSON) |
| 2. Preferences | Set your query pattern (keyword, semantic, hybrid, agentic) and deployment preference |
| 3. Plan | An AI planner designs your search architecture (BM25, semantic, hybrid, or agentic) |
| 4. Execute | Automatically creates OpenSearch indices, ML models, ingest pipelines, and a search UI locally |
| 4.5 Evaluate | (Optional) Evaluate search quality and iterate on the architecture |
| 5. Deploy | (Optional) Deploy to Amazon OpenSearch Service or Amazon OpenSearch Serverless |
After installing, just describe what you want to build:
"Help me build a hybrid search app for my product catalog"
The agent guides you through each phase interactively.
skills/ # All agent skills
opensearch-launchpad/ # Get started with OpenSearch
SKILL.md # Skill instructions (< 500 lines)
scripts/ # Execution scripts
references/ # Loaded on demand per phase
search-relevance/ # Future: query tuning, ranking, evaluation
log-analytics/ # Future: log ingestion, parsing, dashboards
observability/ # Future: traces, metrics, monitoring
kiro/ # Kiro Power integrations
opensearch_orchestrator/ # MCP server (Kiro Power path only)
tests/ # All tests
We welcome new skills! See the Developer Guide for step-by-step instructions on creating a skill, the SKILL.md template, conventions, testing, and the release process.
Phase 5 of opensearch-launchpad deploys your local search solution to AWS. This is optional — Phases 1–4 work entirely locally.
Add these servers to the power's mcp.json configuration in Kiro:
{
"mcpServers": {
"awslabs.aws-api-mcp-server": {
"command": "uvx",
"args": ["awslabs.aws-api-mcp-server@latest"],
"env": { "FASTMCP_LOG_LEVEL": "ERROR" }
},
"aws-knowledge-mcp-server": {
"command": "uvx",
"args": ["fastmcp", "run", "https://knowledge-mcp.global.api.aws"],
"env": { "FASTMCP_LOG_LEVEL": "ERROR" }
},
"opensearch-mcp-server": {
"command": "uvx",
"args": ["opensearch-mcp-server-py@latest"],
"env": { "FASTMCP_LOG_LEVEL": "ERROR" }
}
}
}aws configureOr set environment variables:
export AWS_ACCESS_KEY_ID="your-access-key"
export AWS_SECRET_ACCESS_KEY="your-secret-key"
export AWS_REGION="us-east-1"Your AWS user/role needs permissions for:
- OpenSearch Service — create/manage domains and serverless collections
- IAM — create and manage roles for OpenSearch
- Bedrock — invoke models (for semantic and agentic search)
Some MCP clients cannot find uvx or docker from the JSON config environment.
Fix: Locate the full paths and add them to env.PATH in your MCP config:
which uvx # e.g. /Users/you/.local/bin/uvx
which docker # e.g. /usr/local/bin/dockerThen in Kiro: Cmd+Shift+P → Kiro: Open user MCP config (JSON) and update:
This project is licensed under the Apache License, Version 2.0. See LICENSE.txt for details.
{ "mcpServers": { "opensearch-launchpad": { "command": "uvx", "args": ["opensearch-launchpad@latest"], "env": { "FASTMCP_LOG_LEVEL": "ERROR", "PATH": "/usr/local/bin:/usr/bin:/bin:/Users/you/.local/bin" } } } }