MiniStack
title: MiniStack type: tool created: 2026-05-19 last_updated: 2026-09-06 related: ["Playradar", "radar/tools/Floci", "radar/techniques/LambdaUnitTesting"] sources: ["https://ministack.org/"] radar_quadrant: Tools radar_ring: Assess radar_position: outer
A free, open-source AWS emulator that runs 60+ AWS services locally using real infrastructure components: RDS spins up actual Postgres or MySQL containers, ElastiCache runs a real Redis instance, and ECS executes real Docker containers. MIT licensed with no auth token, no telemetry, and no feature gating.
MiniStack takes a real-infrastructure approach rather than emulating AWS service behaviour in-process. The trade-off: higher fidelity to production behaviour at the cost of heavier resource usage compared to native emulators.
| Metric | MiniStack | LocalStack | Floci |
|---|---|---|---|
| Startup time | ~2 seconds | ~3,300 ms | 24 ms |
| Idle memory | ~30 MiB | 143 MiB | 13 MiB |
| Docker image | ~270 MB | ~1 GB | ~90 MB |
| AWS services | 60+ | varies | 45 |
| Tests passing | 2,400+ | — | 1,925/1,925 |
Works with boto3, the AWS CLI, Terraform, CDK, and other AWS tooling without configuration changes. A Testcontainers Java module is available on Maven Central for integration with Java test suites.
Both MiniStack and Floci emerged as LocalStack alternatives following LocalStack's shift to a paid auth-token model in March 2026. They solve the same problem with different design philosophies:
- Floci (Assess/inner) uses Quarkus Native for 24ms startup and 13MiB idle; optimised for CI/CD cold-start performance.
- MiniStack uses real containers (Postgres, Redis, Docker) for higher production fidelity; better suited to integration tests that depend on real database and cache behaviour.
The choice depends on whether startup speed or infrastructure realism is the higher priority for a given project.
A closer look at per-service depth shows MiniStack's landing page states many services as fully covered (S3, SQS, SNS, DynamoDB, IAM, Route53, EventBridge, Step Functions, and others), but explicitly labels several as stubs with no real behaviour: CloudFront, WAF Classic, CloudTrail (no real event logging), AWS Config (rules not evaluated), Cost & Usage Reports (no real billing data), Backup, and Amazon MQ (no real broker). Unlike Floci's documented SDK-conformance test result, MiniStack's "Full" ratings are self-reported on its own site and have not been independently verified against its issue tracker.
MiniStack sits in the Assess ring of the Tools quadrant, at outer position. First studied via ministack.org on 2026-05-19; no personal use. The real-infrastructure approach is a distinct and legitimate design choice from Floci, warranting its own blip. A direct reward/risk/fidelity comparison against radar/tools/Floci favours Floci: Floci's gaps are vendor-documented and its service fidelity has a checkable SDK-conformance metric, while MiniStack's broader "Full" claims remain unverified marketing text. Position is moved from center to outer to reflect this — MiniStack remains a legitimate choice specifically when a project needs the highest possible fidelity for real-Postgres/Redis-backed integration tests, but is no longer the co-equal alternative to Floci it was at first study. The remaining gate is a project where real Postgres or Redis container behaviour is necessary.
