Comprehensive context for LLMs to build location-aware applications with Amazon Location Service. Provides ready-to-use integrations as a Kiro power, Claude Code plugin, Agent Skill, and direct context files.

This package guides AI coding assistants through adding maps, places search, geocoding, routing, and other geospatial features with Amazon Location Service, including authentication setup, SDK integration, and best practices. It follows progressive disclosure principles—loading minimal context by default and expanding based on your specific task.

Kiro IDE — Install as a Power:

Alternatively, open Kiro IDE → Powers panel → Available tab → search for "Build geospatial applications with Amazon Location Service"

Note: When using Spec mode, include "use the Amazon Location Service power" in your spec prompt for Kiro to activate it.

Kiro CLI — Install as an Agent Skill:

npx skills add aws-geospatial/amazon-location-agent-context -a kiro-cli

After installing, add the skill to your custom agent's resources in .kiro/agents/<agent>.json:

{
  "resources": ["skill://.kiro/skills/**/SKILL.md"]
}

Note: Kiro CLI skill installations don't include MCP configuration automatically. See MCP Servers below for manual setup.

Activate by keywords: Mention "location", "maps", "geocoding", "routing", "places", "geofencing", or "tracking" in your prompts.

Install as a Plugin from the respective official marketplace:

Claude Code:

You can install the amazon-location-service plugin from the official Claude Plugins marketplace:

/plugin install amazon-location-service@claude-plugins-official

Cursor:

You can install the amazon-location-service plugin from the official Cursor Marketplace. For additional information, please refer to the Cursor plugin documentation. You can also install within the Cursor application:

1. Open Cursor Settings
2. Navigate to Plugins
3. Search for AWS
4. Select the amazon-location-service plugin and click Add to Cursor
5. Select the scope for the installed plugin
6. The plugin should appear under Plugins → Installed

Our plugin hasn't been added to the official Codex marketplace yet, so you'll need to install it from the agent-plugins marketplace:

1. Clone the agent-plugins repository locally
2. Open the repository in Codex so it can discover .agents/plugins/marketplace.json
3. Open the plugins with /plugins
4. Navigate to Amazon Location Service
5. Press Enter and choose Install plugin

For Claude Code and Cursor users, we recommend the plugin above for the best experience (includes MCP configuration). For all other agents that support the Agent Skills standard:

The skills/amazon-location-service/ directory is an Agent Skills package — an open standard for giving agents new capabilities. Agent Skills are supported by GitHub Copilot, OpenCode, Codex, Antigravity, and more.

Install with the skills CLI:

npx skills add aws-geospatial/amazon-location-agent-context

The CLI will guide you through selecting which agent(s) to install the skill for and at what scope (project or user level). For example:

$ npx skills add aws-geospatial/amazon-location-agent-context

? Select an agent: (Use arrow keys)
❯ Claude Code
  Cursor
  GitHub Copilot
  OpenCode
  Codex
  Antigravity

? Select a scope: (Use arrow keys)
❯ Project — install in current directory (committed with your project)
  Global — install globally for all projects

Install for a specific agent:

npx skills add aws-geospatial/amazon-location-agent-context -a github-copilot

npx skills add aws-geospatial/amazon-location-agent-context -a opencode

npx skills add aws-geospatial/amazon-location-agent-context -a codex

Once installed, the skill activates automatically when your task involves location, maps, geocoding, routing, or other Amazon Location Service topics. The SKILL.md file includes the required frontmatter (name, description) per the Agent Skills spec, and reference files in references/ are loaded on demand for progressive disclosure.

See agentskills.io for the full list of compatible tools.

If you're not using Kiro, Claude Code/Cursor plugins, or one of the agents supported by Agent Skills, you can load the context files directly into your LLM:

1. Start with context/amazon-location.md for the service overview
2. Add specific files from context/additional/ as needed for your task or can be read as needed by the LLM client

Kiro power and Claude Code/Cursor plugin installations include MCP configuration automatically. If you're using Agent Skills or direct context, configure this server manually for full functionality:

• AWS MCP Server: AWS API exploration, execution, and documentation access
  • Setup instructions

• Progressive Disclosure: In order for these sources of context for LLMs to be generally effective, they must avoid providing context unrelated to the current session. Minimal additional context to guide LLMs through common hallucinations provide a much more usable set of context files.
• Custom Client Integrations: People use LLMs through clients, and those clients have preferred methods for loading context and tooling. We should support preferred client integration methods (Kiro powers, Claude Code plugins, Agent Skills, etc.). The core context files are provided for generality and are tuned in such a way to encourage progressive disclosure for any client, without relying on specific client features.
• Minimal Context Performs Best: LLMs are able to drive a lot of value themselves. In order to effectively leverage this value and avoid adding weight to undesirable results, these context files must address common hallucinations and knowledge gaps but avoid as little extra information as possible.
• Mutating Actions Require Human Approval: All tool configuration in this package will be setup to be as powerful as possible without granting access to mutating actions without additional configuration or manual approval.
• Self-Contained Projections: Each projection (Kiro, Claude, Skills, etc.) must be self-contained for distribution. Content from the src/context/ directory is merged into projections during the build process, as clients only package the projection directory (not the source context files).

The maintainers have some LLM performance benchmarks which are run before publishing updates. We would like to externalize this process, but for now, in order to ensure that additions do not decrease performance on existing supported queries, there is no process for contributions.

All commits updating context information must address why the context is being added to the repo in their commit message. This should include a brief description of the situation that requires the additional steering information and may include why existing steering information in this repo and published documentation is insufficient.

This library is licensed under the MIT-0 License. See the LICENSE file.