AWS Proton is Deprecated: What ECS Teams Must Do Now
Originally published at https://fortem.dev/blog/aws-proton-deprecated
AWS Proton shuts down October 7, 2026. ECS environments keep running but Proton templates and pipelines stop. Here's what breaks and how to migrate.
On October 7, 2026, AWS is shutting down Proton. Your ECS environments will keep running — but the console, the APIs, and every pipeline you built on top of Proton will stop working. This covers what breaks, what doesn't, and how to migrate with enough runway before the deadline.
TL;DR
- October 7, 2026: Proton console goes dark, all Proton API calls return errors — your ECS clusters keep running.
- Any CI/CD pipeline calling aws proton commands breaks on that date; audit CloudTrail now to find them.
- Three replacement paths: Terraform + CI/CD (full control), Fortem (operational layer, running in 7 days, from $790/mo), or a multi-cloud IDP like Humanitec (months to implement).
- Start migration by end of June — parallel-running old and new systems for 4–6 weeks is the only safe cutover strategy.
Deadline — October 7, 2026. After this date, the AWS Proton console becomes inaccessible and all Proton API calls will return errors. You have approximately 4 months from today to migrate.
What AWS Proton was
AWS Proton was a managed service for platform teams — define infrastructure templates once, let developers deploy from them without touching CloudFormation or Terraform.
“AWS Proton automates the provisioning and deployment of serverless and container-based applications.”
— AWS Proton documentation, archived June 2026
AWS Proton launched in 2021 as a managed service for platform teams. The core idea: platform engineers define environment and service templates (in CloudFormation or Terraform), developers use those templates to deploy without needing to understand the underlying infrastructure.
In practice, Proton had a steep learning curve, required owning and maintaining the template library yourself, and provided limited visibility into running environments. Adoption stayed low. Teams that did adopt it often used it as a thin wrapper around existing Terraform or CloudFormation — getting the deployment abstraction without much of the ongoing operations value.
The shutdown isn't a surprise to most Proton users. The product received minimal investment for two years before the deprecation announcement.
What breaks on October 7
The Proton console, API, and any CI/CD pipeline calling Proton APIs stop working. Your running ECS infrastructure keeps operating — only the orchestration layer disappears.
On the deprecation date, AWS will shut down the Proton service layer. This means:
- ✗The AWS Proton console becomes inaccessible
- ✗All Proton API calls (aws proton ...) return errors
- ✗Proton-managed deployment pipelines stop triggering
- ✗Proton environment templates can no longer be used to provision or update environments
- ✗Any automation that calls Proton APIs — GitHub Actions, Lambda functions, scripts — breaks
- ✗Proton service instances can no longer be updated through Proton
What doesn't break
Your ECS clusters, services, task definitions, ALBs, and running containers all survive the October 7 shutdown — Proton managed orchestration, not the underlying compute.
your running infrastructure is not managed by Proton. Proton was an orchestration layer, not the underlying compute. When Proton shuts down:
- ✓Your ECS clusters, services, and task definitions keep running
- ✓Your RDS instances, ElastiCache clusters, and other resources are unaffected
- ✓Your load balancers and networking continue operating normally
- ✓Customer traffic is not interrupted
- ✓Your Terraform state is unchanged
The risk isn't an outage on October 7. The risk is that after October 7, you lose the tooling to manage your environments going forward. Deployments, updates, provisioning new environments — those workflows need to be rebuilt before the deadline, not recovered after it.
The Copilot pattern repeating
AWS deprecated both Copilot (June 12, 2026) and Proton (October 7, 2026) within the same year — a clear signal that AWS is exiting the managed developer-experience layer entirely.
AWS Copilot— another tool for deploying ECS services — had end-of-support announced March 6, 2026 and reaches EOL on June 12, 2026. Both Copilot and Proton served the same category: simplifying ECS deployment and operations for teams that didn't want to write raw CloudFormation or Terraform.
AWS's signal is consistent: they are not in the business of building developer experience tooling. Their investment goes into infrastructure primitives — ECS itself, Fargate, EventBridge, CloudFormation. The operations layer — fleet visibility, scheduling, self-service, environment management — is expected to come from third-party products.
AWS managed ECS tooling — status
AWS CopilotDeprecated June 12, 2026
AWS ProtonDeprecated October 7, 2026
AWS App RunnerActive (single-app PaaS, not fleet management)
Amazon ECS ConsoleActive (infrastructure primitives only)
Your three migration options
Three paths: Terraform modules + CI/CD (full control, high effort), a purpose-built ECS platform like Fortem (faster, no rewrite), or a multi-cloud IDP (broader but heavier).
Raw Terraform + CI/CD pipelines
Replace Proton's template system with Terraform modules for ECS Fargate and deploy via GitHub Actions, AWS CodePipeline, or similar. This is the most common migration path and the right choice if you want full control and have the platform engineering capacity to maintain it.
Advantages
- No new vendor dependency
- Full control over every resource
- Works with any CI/CD system
Trade-offs
- No fleet visibility or scheduling
- Developers still need AWS Console access
- You rebuild the operations layer yourself
Fortem
Fortem is an ECS operations platform built for teams managing 10+ environments. It connects to your existing AWS account via cross-account IAM, picks up your environments automatically, and adds the layer Proton never had: scheduling, fleet-wide visibility, environment cloning, RBAC-scoped self-service, and AI diagnostics. Your Terraform stays unchanged.
Advantages
- No Terraform rewrite required
- Running in 7 business days
- Scheduling saves 60–70% on dev/staging
- Developers self-serve without AWS Console
Trade-offs
- Starts at $790/month
- ECS-first (EKS on roadmap for 2027)
- Managed onboarding, not self-serve signup
General-purpose IDP (Humanitec, Backstage)
Platforms like Humanitec or Backstage can replace what Proton did and more — they're built for multi-cloud, multi-runtime environments where you need a unified developer portal across Kubernetes, ECS, Lambda, and other runtimes.
Advantages
- Multi-cloud and multi-runtime
- Rich plugin ecosystem (Backstage)
- Good for 200+ person engineering orgs
Trade-offs
- Months to implement, not days
- Requires dedicated platform eng team
- Overkill for ECS-only fleets under 200 envs
Migration timeline
You have 4 months until October 7, 2026. Audit what Proton manages, pick a replacement, and run migration in parallel — do not wait for the last week.
You have until October 7 — roughly 4 months from today. The breakdown below is how to use that runway without leaving the cutover to the last week of September.
Now — June
Audit your Proton usage
List every Proton environment template, service template, and pipeline you have. For each one, document what it provisions and what CI/CD workflows depend on it. This is the inventory you'll need regardless of which migration path you choose.
June — July
Choose your replacement and start parallel running
Pick your migration path and start building the replacement alongside existing Proton workflows. Don't cut over in one shot — run both in parallel for 4–6 weeks so you can catch anything Proton was doing that you missed in the audit.
August — September
Migrate all environments, decommission Proton resources
Move all environments to the new system. Delete Proton service instances, environments, and templates — this also stops you paying for any Proton-provisioned resources that are no longer needed. Verify no automation is calling Proton APIs.
Before October 7
Verify zero Proton API calls in logs
Check CloudTrail for any remaining proton:* API calls in the 30 days before the deadline. Anything still calling Proton needs to be updated or it will break silently on October 7.
KEY INSIGHT: You have exactly 4 months until October 7, 2026. Teams that start their Proton migration in June get 4–6 weeks of parallel running; teams that wait until September get 2 weeks. The runway math is fixed — the only variable is when you start.
Fortem connects to your existing AWS account, picks up your ECS environments automatically, and replaces the operational layer Proton never delivered — scheduling, fleet visibility, environment cloning, and RBAC self-service. No Terraform rewrite. Running in 7 business days.
Migration guide for Proton teams: fortem.dev/migrate-from-proton
Top comments (0)