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

Install dependencies:

Create and activate a virtual environment (recommended):

```bash
python3 -m venv .venv
source .venv/bin/activate  # on macOS / Linux
```
Then install the required Python packages:

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

2. 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):

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

```sh
python 1-setup_gateway.py
```

On success it writes gateway_config.json.

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.

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.
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

```json
[
{
    "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/

MCP Servers