Post

Module 11 — Gateway: a .NET Lambda as an agent tool

Posted
By
2 min read

Prompt & Tool Design for .NET Teams · Part 4 — AgentCore and Production · Module 11 of 12

Module 5 exposed a function to the model from inside your own process. Gateway does the same job across the network: it takes a Lambda you already have and presents it to an agent as a tool, deriving the schema for you. No Python, no Model Context Protocol (MCP) server, no new protocol to learn.

Objective

Expose the C# Lambda from Module 8 as a tool via AgentCore Gateway.

Read (~10 min)

Lab (~20 min)

Register the C# Lambda from Module 8 as a Gateway target. Good news since this course was first written: the Amazon Web Services (AWS) provider now ships aws_bedrockagentcore_gateway and aws_bedrockagentcore_gateway_target, so you can declare the whole thing instead of clicking through the console:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

resource "aws_bedrockagentcore_gateway" "example" {
  name     = "comic-gateway"
  role_arn = aws_iam_role.gateway_role.arn

  authorizer_configuration {
    custom_jwt_authorizer {
      discovery_url = "https://your-idp/.well-known/openid-configuration"
    }
  }
}

resource "aws_bedrockagentcore_gateway_target" "comic" {
  name               = "find-comic"
  gateway_identifier = aws_bedrockagentcore_gateway.example.gateway_id
  description        = "Exposes the comic-finder Lambda as an agent tool"

  credential_provider_configuration {
    gateway_iam_role {}
  }

  target_configuration {
    mcp {
      lambda {
        lambda_arn = aws_lambda_function.comic.arn

        tool_schema {
          inline_payload {
            name        = "find_comic"
            description = "Finds a programming or tech comic for a given topic."

            input_schema {
              type = "object"

              property {
                name        = "topic"
                type        = "string"
                description = "Topic or keyword, e.g. git or deadlines"
                required    = true
              }
            }
          }
        }
      }
    }
  }
}

The name and description in that tool schema are the same contract you wrote in Module 3 and implemented in Module 5. Invoke the tool through the Gateway’s endpoint and confirm the schema it presents matches what you intended. If it doesn’t, the Gateway is right and your annotations are wrong — fix the Lambda.

The authorizer above expects a JSON Web Token (JWT) from your identity provider; point discovery_url at your own.

Done when

An agent (or a test harness) calls your .NET code through the Gateway, and you’ve noted anything that still isn’t expressible in Terraform for your Architecture Decision Record (ADR). Provider coverage has caught up a lot, but it’s worth confirming against your pinned version.

Series navigation

Tutorial, AI
artificial-intelligence dotnet cloud
This post is licensed under CC BY 4.0 by the author.