Agentcore Gateway MCP To API

AgentCore Gateway QuickStart 🚀

Amazon Bedrock AgentCore Gateway makes it easy to convert APIs, Lambda functions, and services into MCP-compatible tools and expose them to AI agents. This repo is a minimal quickstart to create a Gateway, register a Lambda target, and run a test agent that invokes the tools.

See the upstream guide for more detail: https://aws.github.io/bedrock-agentcore-starter-toolkit/user-guide/gateway/quickstart.html

Files in this repo

• 1-setup_gateway.py — script to create a Gateway and add a Lambda target. See function setup_gateway.
• 2-run_agent.py — interactive agent tester. See function run_agent.
• lambda_function.py — Lambda implementation exposing tools: lambda_handler, get_weather, get_time.
• gateway_config.json — (generated) saved gateway configuration produced by the setup script.
• discovery-endpoint.json — (generated) OAuth discovery endpoint used by the Gateway.

💰 Cost

This demo will incur a small cost (under $10) to your account, mainly due to Cognito use for machine-to-machine identity. Check the AWS Free Tier, which offers $200 credit for new customers.

Quick start

1. Install dependencies:

Create and activate a virtual environment (recommended):

    python3 -m venv .venv
    source .venv/bin/activate  # on macOS / Linux

Then install the required Python packages:

    pip install boto3
    pip install bedrock-agentcore-starter-toolkit
    pip install strands-agents

1. Run script to create the Gateway and Lambda target:

Note: If you're using a multi-profile AWS configuration, make sure to activate the appropriate profile before running the scripts and setting it up the environment ($AWS_PROFILE):

    export AWS_PROFILE=[a profile from ~/.aws/credentials]

    python 1-setup_gateway.py

• On success it writes gateway_config.json.

1. Update the Lambda code with more functional one.

✳️ Look up in the account the Lambda function created and replace with provided code, over the initial boilerplate.

1. Enjoy it in the interactive agent chat:

    python 2-run_agent.py

- This runs run_agent. - The agent uses the Bedrock model (change model_id inside the script as needed) and lists available tools from the Gateway.

1. Prompt ideas and poke the code

- Ask the agent about weather or time: Try time in Paris, Time in France, then time in some obscure vilage. Try in another language. Enjoy the flexability of this kide of user facing interface, against a minimal API backend code. - Check the Lambda tools as defined in lambda_function.py, also see the MCP definition below. - Lambda tools: - get_weather: accepts location - get_time: accepts timezone

Notes & customization

• Change region in 1-setup_gateway.py before running if you are not in eu-west-1.
• Update model in 2-run_agent.py by editing model_id.
• Replace or extend the Lambda implementation in lambda_function.py to connect to other services. (update via Lambda console)
• The setup uses Cognito and generates client details to a file gateway_config.json.
• OAuth discovery details are generated and stored in discovery-endpoint.json.

MCP definition for reference

or view/edit in the AWS console

    [
    {
        "description": "Get weather for a location",
        "inputSchema": {
        "properties": {
            "location": {
            "type": "string"
            }
        },
        "required": [
            "location"
        ],
        "type": "object"
        },
        "name": "get_weather"
    },
    {
        "description": "Get time for a timezone",
        "inputSchema": {
        "properties": {
            "timezone": {
            "type": "string"
            }
        },
        "required": [
            "timezone"
        ],
        "type": "object"
        },
        "name": "get_time"
    }
    ]

Troubleshooting

• If gateway_config.json is missing, run setup_gateway again.
• If the agent cannot fetch tools, check access token retrieval in run_agent and the values in gateway_config.json.

License & references

• This quickstart is a minimal example. For full operations and advanced configuration, see the AgentCore Starter Toolkit documentation:

https://aws.github.io/bedrock-agentcore-starter-toolkit/