MoEngage turned 400 private APIs into agent tools with AgentCore Gateway, saving six months of engineering
AWS ML Blog · rss · 2026-10-30
An AWS ML Blog deep dive on how MoEngage made 400 internal REST APIs (25 teams, 7 Regions, VPC-private) safely callable by AI agents using Amazon Bedrock AgentCore Gateway, saving an estimated six months of platform engineering and cutting API onboarding from days to hours.
Core principle: teams don't build MCP servers—each just ships an OpenAPI spec; the AI platform team owns the MCP layer end to end.
Four constraints: no public endpoints ever; identity/tenant context injected server-side, never exposed as model-settable parameters; and server-side tool filtering, since a full 400-tool catalog would cost 20,000+ tokens per connection—client-side filtering isn't an authorization boundary.
How the gateway maps to these: managed MCP server generates tools directly from OpenAPI specs (TargetoperationId naming); Cedar policies with partial evaluation filter tools/list per principal before serialization (e.g., a QA bot never even receives write tools); managed egress over VPC Lattice removes per-API networking; built-in semantic search lets agents find tools by task description, also eliminating rename fragility.
More from coding & agent
- Scout24's CTO rebuilds org-wide agent workflows on Linear, from idea to merged PR — petergyang · 2026-10-30
- Henge: Open-Source ComfyUI Video Editing Interface Built on Custom MiniMax H3 LoRAs — Round_Awareness5490 · 2026-10-30
- Lean 4 formalization of an asymmetric metric-tensor physics paper hits GitHub — SubtleGradient · 2026-10-30
- YC-backed agent-native office suite launches free in ChatGPT and Codex — ycombinator · 2026-10-30
- Supabase never had an agent strategy — its CTO explains why it won anyway — socialwithaayan · 2026-10-30
- Understand Anything hits 88.9k stars: turn any codebase into a queryable knowledge graph — adnan_hashmi · 2026-10-30