{"id":94749,"date":"2026-10-16T15:14:01","date_gmt":"2026-10-16T13:14:01","guid":{"rendered":"https:\/\/www.skaylink.com\/?p=94749"},"modified":"2026-10-22T10:02:47","modified_gmt":"2026-10-22T08:02:47","slug":"bedrock-agents-classic-agentcore-mantle-difference","status":"publish","type":"post","link":"https:\/\/www.skaylink.com\/en\/insights\/blog\/bedrock-agents-classic-agentcore-mantle-difference\/","title":{"rendered":"Bedrock Agents Classic, AgentCore, and Mantle: What\u2019s the difference?"},"content":{"rendered":"\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

Home<\/a><\/span><\/span><\/p>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t

\n\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\"Mann\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Blog<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

Bedrock Agents Classic, AgentCore, and Mantle: What\u2019s the difference?<\/h1>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\tLearn how Amazon Bedrock AgentCore differs from Agents Classic and Bedrock Mantle, and what the migration means for agent architecture and AI inference.\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t
\n\t\t\t\t\t
\n\t\t
\n\t\t\t\t\t
\n\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\tOctober 16, 2026\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t
\n\t\t\t\t\t
\n\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\t\t\"Picture\n\t\t\t\t<\/div>\n\t\t\t\n\t\t\t
\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t\n\t\t\t\t\t\t\tMatina Lysikatou\t\t\t\t\t\t<\/span>\n\t\t\t\t\t<\/div>\n\t\t\t\t\n\t\t\t\t\t\t\t\t\t
\n\t\t\t\t\t\t

Senior AWS Consultant<\/p>\n\t\t\t\t\t<\/div>\n\t\t\t\t\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t

\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\n\t\t\t
\n\t\t\t\t\n\t\t\t\t\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t
\n\t\t\t\t\t
\n\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

News<\/strong>: Bedrock Agents, renamed Bedrock Agents Classic, has been unavailable to new customers\/accounts since July 30, 2026. Existing accounts can continue using the service in maintenance mode. AWS recommends Amazon Bedrock AgentCore as the successor. Please refer to:\u00a0Amazon Bedrock Agents Classic maintenance mode \u2013 Amazon Bedrock<\/span><\/span><\/a><\/span><\/span><\/span><\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

If you currently use or are planning to use AWS Bedrock Agents as your AI platform, you should be aware of this issue.<\/p>

A possible approach is to upgrade to AgentCore as the runtime and then use LangGraph or Strands as the framework. AgentCore is framework-agnostic, which is a significant advantage.<\/p>

Bedrock Agents Classic<\/strong> was a managed, configuration-based agent orchestrator (fixed loop, Action Groups, Knowledge Bases). AgentCore<\/strong>\u00a0represents a different paradigm: It is a bring-your-own-agent (BYOA) infrastructure platform. You can build your agent using a framework of your choice (LangGraph, CrewAI, Strands, Google ADK, raw code).\u00a0 AgentCore provides the underlying infrastructure, including session isolation, memory, tool and gateway connectivity, identity and authentication observability, a code interpreter, and a browser tool. It is described as a \u201cruntime substrate,\u201c not an agent orchestrator itself.<\/p>

AgentCore<\/strong> is Amazon\u2019s new agent infrastructure layer. For the actual LLM calls, AgentCore is explicitly model-agnostic<\/u>: It works with Amazon Bedrock models (via bedrock-mantle or bedrock-runtime), OpenAI, Google Gemini or any LiteLLM-compatible provider. Agents can even switch between model providers during a session. As an agent-hosting layer, AgentCore has three distinct service endpoints and sits above model inference. It can also route requests to several providers:<\/p>
AgentCore surface<\/strong><\/h6><\/td>
Endpoint pattern<\/strong><\/h6><\/td><\/tr>

Control plane (create\/manage runtimes, endpoints, gateways)<\/strong><\/p><\/td>

bedrock-agentcore-control.{region}.amazonaws.com<\/p><\/td><\/tr>

Data plane (invoke a deployed agent runtime)<\/strong><\/p><\/td>

bedrock-agentcore.{region}.amazonaws.com<\/p><\/td><\/tr>

Gateway (tool\/MCP connectivity)<\/strong><\/p><\/td>

{gatewayId}.gateway.bedrock-agentcore.{region}.amazonaws.com<\/p><\/td><\/tr><\/tbody><\/table>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Bedrock Mantle<\/strong> is a separate, lower-level inference engine within Amazon Bedrock that AgentCore can call when using a Bedrock model.  AgentCore itself neither owns or replaces it. Mantle provides a newer OpenAI- and Anthropic-compatible endpoint surface with Zero Operator Access (ZOA) and Zero Data Retention (ZDR) enabled by default.<\/p>\n

If your agent uses a Bedrock-hosted model as its LLM, that specific model call<\/em>  goes through bedrock-runtime (or bedrock-mantle for a Mantle-exclusive model). However, this is an outbound dependence of AgentCore, not AgentCore\u2019s own API surface.<\/p>\n

In a nutshell, here\u2019s what we learned: Migrating from Bedrock Agents Classic to AgentCore is a viable option, but bear in mind that AgentCore is an agent-hosting\/orchestration platform with its own dedicated endpoints, not an inference engine. It does not use bedrock-runtime or bedrock-mantle as its interface. Instead, it sits a layer above both and can access either one, as well as non-Bedrock LLM providers.<\/p>\n

There are two common ways to use AgentCore<\/strong>:<\/p>\n\n\n\n\n\n\n\n
\n
Capability<\/strong><\/h6>\n<\/td>\n
\n
AgentCore harness<\/strong><\/h6>\n<\/td>\n
\n
AgentCore Runtime (BYOA)<\/strong><\/h6>\n<\/td>\n<\/tr>\n
\n

What it is<\/strong><\/p>\n<\/td>\n

\n

Managed orchestration loop (configuration-based, powered by Strands Agents)<\/p>\n<\/td>\n

\n

Serverless hosting for agent code you write in any framework<\/p>\n<\/td>\n<\/tr>\n

\n

Model selection<\/strong><\/p>\n<\/td>\n

\n

\u2705 Configuration field, no code<\/p>\n<\/td>\n

\n

\ud83d\udd35 Custom code required<\/p>\n<\/td>\n<\/tr>\n

\n

Switch model
mid-session<\/strong><\/p>\n<\/td>\n

\n

\u2705 Built-in<\/p>\n<\/td>\n

\n

\ud83d\udd35 Custom code required<\/p>\n<\/td>\n<\/tr>\n

\n

Agent framework choice<\/strong><\/p>\n<\/td>\n

\n

\u274c Fixed (Strands-based)<\/p>\n<\/td>\n

\n

\ud83d\udd35 Any framework (LangGraph, CrewAI, Strands, raw code)<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t

\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Further reading: AgentCore harness vs. Runtime – Amazon Bedrock AgentCore<\/a><\/p>\n\n\n\n\n\n\n\n\n\n\n\n
\n
\n
Option<\/span><\/strong><\/h6>\n
\n
\n
 <\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/th>\n
\n

What it is<\/span><\/strong><\/p>\n

\n
\n
\n
 <\/div>\n<\/div>\n<\/div>\n<\/figure>\n<\/th>\n
\n

Model control plane<\/span><\/strong><\/p>\n

\n
\n
\n
 <\/div>\n<\/div>\n<\/div>\n<\/figure>\n<\/th>\n
\n

Residency behavior<\/span><\/strong><\/p>\n

\n
\n
\n
 <\/div>\n<\/div>\n<\/div>\n<\/figure>\n<\/th>\n
\n

Best for<\/span><\/strong><\/p>\n

\n
\n
\n
 <\/div>\n<\/div>\n<\/div>\n<\/figure>\n<\/th>\n<\/tr>\n
\n
bedrock-runtime<\/code><\/h6>\n<\/th>\n
\n
AWS-native + OpenAI\/Anthropic-compatible endpoint on Bedrock’s Mantle engine<\/h6>\n<\/td>\n
\n
Bedrock-managed FMs + cross-Region profiles<\/h6>\n<\/td>\n
\n
In-Region by default; geoographic\/global profiles optional. Provides Guardrails, prompt caching, intelligent routing, and more granular IAM<\/h6>\n<\/td>\n
\n
General-purpose production inference, Guardrails, prompt routing.<\/h6>\n
Only bedrock-runtime supports geographic\/\/EU\/global cross-Region at all \u2014 cross-Region inference profiles (geographic or global). Mantle does not support cross-Region routing; it\u2019s always tied to the Region of the endpoint being called.<\/h6>\n<\/td>\n<\/tr>\n
\n
bedrock-mantle<\/code><\/h6>\n<\/th>\n
\n
OpenAI\/Anthropic-compatible  endpoint on Bedrock’s Mantle engine<\/h6>\n<\/td>\n
\n
Bedrock-managed FMs, broader model catalog and earlier access to new models<\/h6>\n<\/td>\n
\n
Always in-Region per deployed endpoint (no cross-Region routing).<\/h6>\n<\/td>\n
\n
Agentic tool use, asynchronous jobs, Mantle-exclusive models (e.g., Gemma 4)<\/h6>\n
Use it instead of bedrock-runtime only if you need a Mantle-exclusive capability (web search tool, asynchronous\/background jobs, Projects, or a Mantle-only model like Gemma 4)<\/h6>\n
Use it for EU Sovereign Cloud use cases, where residency in ESC is enforced structurally (separate partition, single Region, zero operator access) rather than through a geograhic or global profile choice, since ESC does not support cross-Region inference.<\/h6>\n<\/td>\n<\/tr>\n
\n
Bedrock Marketplace<\/strong><\/h6>\n<\/th>\n
\n
Catalog of 100+ third-party\/open models (Hugging Face, specialized vendors) deployed via Bedrock<\/h6>\n<\/td>\n
\n
Bedrock-managed, some via managed endpoints that you provision<\/h6>\n<\/td>\n
\n
Region-dependent per model listing<\/h6>\n<\/td>\n
\n
Access to niche\/open models without leaving the Bedrock API surface<\/h6>\n<\/td>\n<\/tr>\n
\n
Bedrock Custom Model Import<\/strong><\/h6>\n<\/th>\n
\n
Bring your own fine-tuned open-weight model (e.g., a SageMaker-trained Llama\/Mistral\/Nova) into Bedrock for serverless inference<\/h6>\n<\/td>\n
\n
You own the weights; Bedrock serves them<\/h6>\n<\/td>\n
\n
Same Region controls as native Bedrock models<\/h6>\n<\/td>\n
\n
Custom\/fine-tuned models without managing the serving infrastructure<\/h6>\n<\/td>\n<\/tr>\n
\n
SageMaker AI real-time endpoints<\/strong><\/h6>\n<\/th>\n
\n
Dedicated, self-managed inference endpoints (your choice of instance type, container)<\/h6>\n<\/td>\n
\n
Full control \u2014 any model, any container<\/h6>\n<\/td>\n
\n
You pick the exact Region\/Availability Zone; no AWS-managed cross-Region routing<\/h6>\n<\/td>\n
\n
Latency\/cost\/throughput tuning, models not available on Bedrock<\/h6>\n<\/td>\n<\/tr>\n
\n
SageMaker HyperPod<\/strong><\/h6>\n<\/th>\n
\n
Resilient, large-scale training\/inference infrastructure on EKS or Slurm<\/h6>\n<\/td>\n
\n
Full infrastructure control, orchestration layer exposed<\/h6>\n<\/td>\n
\n
Entirely within whatever Region\/VPC you deploy<\/h6>\n<\/td>\n
\n
Large-scale custom training and inference at scale, maximum control<\/h6>\n<\/td>\n<\/tr>\n
\n
Self-hosted (EC2\/EKS + vLLM\/TGI\/Triton)<\/strong><\/h6>\n<\/th>\n
\n
You deploy and operate the serving stack yourself<\/h6>\n<\/td>\n
\n
Full control<\/h6>\n<\/td>\n
\n
Entirely your choice \u2014 single Availability Zone, single Region, or multi-Region<\/h6>\n<\/td>\n
\n
Workloads requiring full infrastructure ownership, air-gapped or heavily audited environments<\/h6>\n<\/td>\n<\/tr>\n
\n
AgentCore
 (harness or Runtime)<\/strong><\/h6>\n<\/th>\n
\n
Agent orchestration\/hosting layer, calls into any of the above as its model backend<\/h6>\n<\/td>\n
\n
N\/A \u2014 it’s a consumer of inference, not a provider<\/h6>\n<\/td>\n
\n
Inherits residency from whichever inference layer + Region you point it at<\/h6>\n<\/td>\n
\n
Production agents needing session isolation, memory, tool gateway, identity, observability<\/h6>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Planning or working on an architecture for an EU-sovereign use case? Stay tuned for our next post.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"

Learn how Amazon Bedrock AgentCore differs from Agents Classic and Bedrock Mantle, and what the migration means for agent architecture and AI inference.<\/p>\n","protected":false},"author":23,"featured_media":94758,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[100],"tags":[182,168],"class_list":["post-94749","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-blog","tag-ai-ml-2","tag-aws"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/posts\/94749","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/users\/23"}],"replies":[{"embeddable":true,"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/comments?post=94749"}],"version-history":[{"count":32,"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/posts\/94749\/revisions"}],"predecessor-version":[{"id":95233,"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/posts\/94749\/revisions\/95233"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/media\/94758"}],"wp:attachment":[{"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/media?parent=94749"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/categories?post=94749"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.skaylink.com\/en\/wp-json\/wp\/v2\/tags?post=94749"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}