From 99e9adf6ee67f076fee3df599b39bfab368e0961 Mon Sep 17 00:00:00 2001 From: Logan Kleier Date: Wed, 2 Sep 2026 10:12:49 -0700 Subject: [PATCH 1/3] fix(architect): recommend AgentCore over classic Bedrock Agents Classic Bedrock Agents (bedrock-agent) entered maintenance mode and closed to new customers on 2026-07-30, so it can no longer be adopted for new builds. Update the architect-for-startups Strands decision guide to recommend AgentCore for managed multi-agent orchestration, matching the gcp-to-aws migration skill and AWS guidance. Refs: - https://docs.aws.amazon.com/bedrock/latest/userguide/agents-classic-maintenance-mode.html - https://aws.amazon.com/about-aws/whats-new/2026/06/aws-service-availability/ --- .../references/strands-agent.md | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md index 160755282..f5085b87f 100644 --- a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md +++ b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md @@ -2,12 +2,12 @@ ## When to Use Strands (vs. Alternatives) -| Situation | Recommendation | Why | -| -------------------------------------------- | -------------------- | --------------------------------------------------------------------------- | -| Building first AI agent on AWS | **Strands** | Thinnest abstraction, least vendor lock-in, direct Bedrock integration | -| Already invested in LangChain/LangGraph | Stay on LangChain | Migration cost isn't worth it unless you're hitting LangChain-specific pain | -| Need managed multi-agent orchestration | Bedrock Agents | If you don't want to manage containers and agent routing yourself | -| Simple single-call LLM feature (no tool use) | Direct `InvokeModel` | Strands adds overhead you don't need for prompt-in/text-out | +| Situation | Recommendation | Why | +| -------------------------------------------- | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | +| Building first AI agent on AWS | **Strands** | Thinnest abstraction, least vendor lock-in, direct Bedrock integration | +| Already invested in LangChain/LangGraph | Stay on LangChain | Migration cost isn't worth it unless you're hitting LangChain-specific pain | +| Need managed multi-agent orchestration | AgentCore | Managed runtime/memory/gateway without managing containers or agent routing yourself. (Classic Bedrock Agents is in maintenance mode and closed to new customers as of 2026-07-30 — not an option for new builds.) | +| Simple single-call LLM feature (no tool use) | Direct `InvokeModel` | Strands adds overhead you don't need for prompt-in/text-out | ## TypeScript vs Python: Startup Perspective From 00837ad3ceffe83dc477c7e456d2db82a3a555ec Mon Sep 17 00:00:00 2001 From: Logan Kleier Date: Wed, 2 Sep 2026 19:08:23 -0700 Subject: [PATCH 2/3] fix(architect): close A2A/AgentCore multi-agent gap flagged in review agentcore.md:49 still pointed at classic Bedrock Agents via the 'Bedrock Multi-Agent' phrase, which the earlier grep sweep missed because it doesn't match 'bedrock-agent' or 'Bedrock Agents'. Both files now name a concrete managed-orchestration answer (Strands multi-agent primitives on AgentCore Runtime, or A2A for cross-org) instead of leaving the gap the reviewer called out. --- .../architect-for-startups/references/agentcore.md | 10 +++++----- .../references/strands-agent.md | 12 ++++++------ 2 files changed, 11 insertions(+), 11 deletions(-) diff --git a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/agentcore.md b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/agentcore.md index 2748b84ac..4654937f7 100644 --- a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/agentcore.md +++ b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/agentcore.md @@ -42,11 +42,11 @@ Month 6+: Add Memory, Gateway, Identity as specific needs emerge ### Multi-Agent: Don't Go There Early -| Team size | Agent architecture | Why | -| ------------- | ------------------------------------------- | -------------------------------------------------------------------- | -| 1-3 engineers | Single agent, direct Bedrock calls | You can't debug multi-agent orchestration AND build product features | -| 4-8 engineers | One supervisor + 2-3 specialist agents max | Complexity grows exponentially with agent count | -| 8+ engineers | Multi-agent with A2A or Bedrock Multi-Agent | You have the team to own the operational complexity | +| Team size | Agent architecture | Why | +| ------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | +| 1-3 engineers | Single agent, direct Bedrock calls | You can't debug multi-agent orchestration AND build product features | +| 4-8 engineers | One supervisor + 2-3 specialist agents max | Complexity grows exponentially with agent count | +| 8+ engineers | Multi-agent via Strands (A2A for cross-org, or a Strands supervisor on AgentCore Runtime) | You have the team to own the operational complexity — classic Bedrock Agents' built-in multi-agent collaboration is in maintenance mode and closed to new customers as of 2026-07-30 | ### PoC to Production Pitfall diff --git a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md index f5085b87f..b40c63106 100644 --- a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md +++ b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md @@ -2,12 +2,12 @@ ## When to Use Strands (vs. Alternatives) -| Situation | Recommendation | Why | -| -------------------------------------------- | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -| Building first AI agent on AWS | **Strands** | Thinnest abstraction, least vendor lock-in, direct Bedrock integration | -| Already invested in LangChain/LangGraph | Stay on LangChain | Migration cost isn't worth it unless you're hitting LangChain-specific pain | -| Need managed multi-agent orchestration | AgentCore | Managed runtime/memory/gateway without managing containers or agent routing yourself. (Classic Bedrock Agents is in maintenance mode and closed to new customers as of 2026-07-30 — not an option for new builds.) | -| Simple single-call LLM feature (no tool use) | Direct `InvokeModel` | Strands adds overhead you don't need for prompt-in/text-out | +| Situation | Recommendation | Why | +| -------------------------------------------- | ----------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | +| Building first AI agent on AWS | **Strands** | Thinnest abstraction, least vendor lock-in, direct Bedrock integration | +| Already invested in LangChain/LangGraph | Stay on LangChain | Migration cost isn't worth it unless you're hitting LangChain-specific pain | +| Need managed multi-agent orchestration | Strands multi-agent primitives (Agents-as-Tools, Swarms, Graphs, or A2A for cross-org) on AgentCore Runtime | AgentCore gives you a managed runtime/memory/gateway so you're not managing containers — but you still own the agent routing logic yourself via Strands. (Classic Bedrock Agents' built-in multi-agent collaboration is in maintenance mode and closed to new customers as of 2026-07-30 — not an option for new builds.) | +| Simple single-call LLM feature (no tool use) | Direct `InvokeModel` | Strands adds overhead you don't need for prompt-in/text-out | ## TypeScript vs Python: Startup Perspective From 8b14169c3e8afd0e46e28862615f7b968e5e3bfa Mon Sep 17 00:00:00 2001 From: Logan Kleier Date: Thu, 17 Sep 2026 14:44:31 -0700 Subject: [PATCH 3/3] docs(architect): clarify AgentCore deployment ownership Signed-off-by: Logan Kleier Co-authored-by: Cursor --- .../references/strands-agent.md | 31 ++++++++++++------- 1 file changed, 19 insertions(+), 12 deletions(-) diff --git a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md index b40c63106..17c6e6e05 100644 --- a/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md +++ b/advisor/plugins/aws-startup-advisor/skills/architect-for-startups/references/strands-agent.md @@ -2,12 +2,12 @@ ## When to Use Strands (vs. Alternatives) -| Situation | Recommendation | Why | -| -------------------------------------------- | ----------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -| Building first AI agent on AWS | **Strands** | Thinnest abstraction, least vendor lock-in, direct Bedrock integration | -| Already invested in LangChain/LangGraph | Stay on LangChain | Migration cost isn't worth it unless you're hitting LangChain-specific pain | -| Need managed multi-agent orchestration | Strands multi-agent primitives (Agents-as-Tools, Swarms, Graphs, or A2A for cross-org) on AgentCore Runtime | AgentCore gives you a managed runtime/memory/gateway so you're not managing containers — but you still own the agent routing logic yourself via Strands. (Classic Bedrock Agents' built-in multi-agent collaboration is in maintenance mode and closed to new customers as of 2026-07-30 — not an option for new builds.) | -| Simple single-call LLM feature (no tool use) | Direct `InvokeModel` | Strands adds overhead you don't need for prompt-in/text-out | +| Situation | Recommendation | Why | +| -------------------------------------------- | ----------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | +| Building first AI agent on AWS | **Strands** | Thinnest abstraction, least vendor lock-in, direct Bedrock integration | +| Already invested in LangChain/LangGraph | Stay on LangChain | Migration cost isn't worth it unless you're hitting LangChain-specific pain | +| Need multi-agent orchestration | Strands multi-agent primitives (Agents-as-Tools, Swarms, Graphs, or A2A for cross-org) on AgentCore Runtime | AgentCore manages runtime infrastructure and can provide memory/gateway/identity, but you still own routing via Strands. Direct code deployment avoids image management; with container deployment, you own image maintenance. Classic Bedrock Agents' built-in collaboration is in maintenance mode and closed to new customers as of July 30, 2026. | +| Simple single-call LLM feature (no tool use) | Direct `InvokeModel` | Strands adds overhead you don't need for prompt-in/text-out | ## TypeScript vs Python: Startup Perspective @@ -42,15 +42,22 @@ Memory modes ranked by startup relevance: **Cost of premature LTM**: Memory extraction runs additional model calls per session. At 1000 sessions/day, that's meaningful token spend for personalization most early users won't notice. -## Deployment: The Container Gotcha (TypeScript) +## Deployment: Direct Code vs. Containers (TypeScript) -TypeScript agents REQUIRE containerized deployment (`--deployment-type container`). This means: +AgentCore Runtime supports [direct code deployment for Node.js, including compiled +TypeScript](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/runtime-get-started-code-deploy-node.html). +Prefer it when the supported runtime fits; it avoids owning a container image. -- ECR image build in your CI/CD pipeline -- Container image maintenance (base image updates, dependency patches) -- Slightly higher cold-start than Python agents +Choose container deployment when you need a custom runtime, native dependencies, +or image-level control. Under the [container shared-responsibility model](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/runtime-get-started-code-deploy.html), +you own: -**If you're deploying to Lambda for cost reasons (scale-to-zero)**: Use Python Strands agents — they work with Lambda's native runtime. TypeScript agents need Lambda container image support (slower cold starts, 10GB image limit). +- ECR image builds and CI/CD +- Base-image and dependency patching +- Rebuilding and redeploying images for security updates + +AgentCore manages the runtime infrastructure in either mode; it does not take over +your Strands supervisor/routing logic. ## Evaluation: Ship Evals from Day One (But Cheaply)