Artificial intelligence news.
Straight from the original announcements.

MoEngage turns 400 private APIs into agent tools with AgentCore Gateway

October 29, 2026, 17:54By the AI News Bites Editorial TeamSource: AWS
MoEngage turns 400 private APIs into agent tools with AgentCore Gateway
Image: AWS

MoEngage, the customer engagement platform used by more than 1,350 consumer brands, has used Amazon Bedrock AgentCore Gateway to turn roughly 400 internal REST APIs into permission-scoped tools for AI agents. According to the company's account on the AWS blog, the approach saved an estimated six months of platform engineering effort.

Key points
  • MoEngage exposed about 400 internal REST APIs, built by 25 teams across seven AWS Regions, as agent tools via AgentCore Gateway.
  • Cedar policies filter the tool list per agent on the server side, so agents never receive tools they cannot use.
  • Managed egress over Amazon VPC Lattice and a VPC interface endpoint mean no public endpoint exists in the path.
  • MoEngage estimates it saved six months of platform engineering, and onboarding a new API now takes hours instead of days.

The problem: hundreds of private APIs, no public exposure

The capabilities MoEngage's agents need, such as campaign creation, segment queries and delivery stats, already exist as internal APIs. Behind the product sit roughly 400 internal REST APIs, built by 25 engineering teams and running in private virtual private clouds (VPCs) across seven AWS Regions. MoEngage isolates them regionally to meet data residency commitments and does not expose them to the public internet.

The hard part, according to the post, was making these APIs reachable by agents safely. The post was co-written with Nikhil Kumar and Shakir Khan from MoEngage.

The obvious route was for each team to run its own Model Context Protocol (MCP) server. MoEngage costed that out and rejected it. Over 20 MCP servers would mean over 20 deployments to monitor, patch and keep protocol-compatible, and cross-domain agents would have to connect to several servers and reconcile them. Every team would also need MCP expertise unrelated to its product work.

Instead, MoEngage set a principle: individual engineering teams do not implement MCP. Teams build REST APIs and declare which operations should be tools, and the AI Platform team owns the MCP layer. A team's responsibility ends at a well-described OpenAPI spec.

Four constraints behind the design

MoEngage defined four constraints that all had to be met at once:

  • No public endpoints. The internal APIs are VPC-private by policy. Per-integration networking such as NAT gateways, reverse proxies or VPN tunnels, multiplied across seven Regions, would grow the operational surface faster than the set of agents.
  • Identity is not the agent's decision. Every internal API call carries tenant, user, role and database context. If that were a visible tool parameter, a model could get it wrong or be talked into changing it. Identity had to be injected by the calling service and kept out of the model's reach.
  • The catalog must fit in a context window. At roughly 50 tokens of JSON schema per tool, a full tools/list across the estate is over 20,000 tokens on every connection. Client-side filtering does not fix this, since the full catalog has already been transmitted and is not an authorization boundary.
  • No custom central server. A custom MCP server would still require a tool-generation pipeline, a policy file format, a tool-alias registry for renames and circuit breakers, all owned by one team.

How AgentCore Gateway fits

Amazon Bedrock AgentCore Gateway is described as a fully managed MCP server. Teams register OpenAPI specs as targets, and the gateway generates MCP tools from them and routes calls to the backing services. Each operation becomes a tool named Target___operationId, for example CampaignsAPI___searchCampaigns. When a spec changes, the tool changes, with no deployment or coordination with the AI Platform team.

Four properties mapped onto MoEngage's constraints:

  • Cedar policies. The AgentCore policy engine uses partial evaluation to filter tools/list per principal. Each agent authenticates as its own IAM role, and its Cedar policy decides which tools it can see and call. A campaign QA bot, for instance, gets read access to campaign and segment tools but cannot call write operations, and it never receives write tools in its tools/list response.
  • Managed egress over Amazon VPC Lattice. The gateway reaches internal services through AgentCore managed egress. Inbound, it is reachable only through a VPC interface endpoint with private DNS. Onboarding a new API adds a target, not a network topology change.
  • Semantic search. With the x_amz_bedrock_agentcore_search capability, an agent can find tools by describing the task. It also removes rename fragility: if get_segment_user_count becomes get_segment_size, a search for segment user count still finds it.
  • OpenAPI in, MCP tools out. The spec teams already maintain serves as the tool definition.

In the request flow, an agent service on Amazon EC2 or in a container calls an MCP client, which can be the official MCP SDK, LangChain, Google ADK or MoEngage's own agent runner. The client signs the request with SigV4 using the service's existing IAM role and sends it over HTTPS to the VPC interface endpoint. The gateway then evaluates Cedar policies to determine what the principal can see or call.

What it means

According to the post, no product team wrote MCP server code, no networking was provisioned per integration, and no public endpoint exists anywhere in the path. Onboarding a new API now takes hours instead of days, and the effort saved is estimated at six months of platform engineering.

For other teams with large estates of private APIs, the case study shows one way to expose them to agents while keeping authorization on the server side and the tool list small. These figures are MoEngage's own estimates as reported by AWS.

Related news

AWS details four ways to speed up Amazon Bedrock AgentCore sessions
For developersNovember 13, 2026, 23:02

AWS details four ways to speed up Amazon Bedrock AgentCore sessions

AWS has outlined four architecture patterns for reducing session startup delays in Amazon Bedrock AgentCore runtime, including snapshot-based initialization, build hooks…

AWS upgrades AgentCore runtime with faster starts and stronger isolation
Products and toolsNovember 12, 2026, 18:59

AWS upgrades AgentCore runtime with faster starts and stronger isolation

AWS has announced updates to the Amazon Bedrock AgentCore runtime, giving AI agent sessions faster startup, more compute capacity and expanded isolation for data. The…

AWS launches direct Web Search API for Bedrock AgentCore
Products and toolsNovember 12, 2026, 00:13

AWS launches direct Web Search API for Bedrock AgentCore

AWS has introduced the Web Search API on Amazon Bedrock AgentCore, allowing applications to retrieve ranked web results through a single API call. The service is…