macrostack
Browse

The AI stack

Categories

Local & Sovereign AINotes & KnowledgeObservability & MonitoringPassword ManagersWeb AnalyticsTeam ChatSmart HomeNetworking & RoutersVideo ConferencingCloud Storage & SyncPhotos & MediaAPI DevelopmentImage EditingWorkflow Automation & iPaaSDeveloper Tools & ContainersOffice & Productivity SuitesNo-Code DatabasesCode Hosting & Git ForgesProject ManagementEmail Marketing & NewslettersScheduling & BookingError Tracking & Exception MonitoringLog Management & SIEMVPN & PrivacyEmail & Secure MailVector Databases & AI SearchLLM & Agent FrameworksDomains & Web HostingData Removal & PrivacyAuthentication & IdentityHelp Desk & Customer SupportCloud & VPSKubernetes & Container PlatformsEmbedding ModelsPDF & DocumentsAI Coding AssistantsAI Voice & SpeechLLM Observability & EvaluationLLM Gateways & RoutingCloud GPU & AI ComputeCI/CD & build automationData & pipeline orchestrationModel serving & inferenceAI agent frameworksBackend as a serviceSecrets managementFeature flags & experimentationProduct analyticsSearch infrastructureUptime & status monitoringAffiliate & partner platformsVisitor identification & personalisationWikis & internal docsIdentity & access managementData warehouses & analytics enginesCustomer data platformsCRMObject storageBI & dashboardsE-signatureWhiteboards & diagrammingIn-memory data stores & cachingPlatform as a serviceTransactional & bulk emailHeadless CMSDesign & prototypingE-commerce platformsInternal tools & admin panelsManaged databasesForms & surveysFine-Tuning & Model TrainingRAG & Retrieval PlatformsLLM Evaluation & TestingAI Guardrails & Content SafetySpeech Recognition & TranscriptionExperiment Tracking & ML OpsDocument AI & OCR

About

How we rank & score

Step by step

Migration guides

65 moves, each one written from the real export and import paths. What it costs, what carries over, what you rebuild by hand — and the conditions under which staying where you are is the better call.

  • AlgoliaMeilisearch

    Algolia's Grow tier runs about $1 per 1,000 search requests plus $0.50 per 1,000 records, and search-as-you-type fires a request per keystroke. A site doing 500,000 monthly searches pays roughly $500; Meilisearch on a $40 VPS costs $40.

    effort: One to two weeks including relevance tuning.all Algolia alternatives →
  • AlgoliaTypesense

    Same Algolia baseline — roughly $500/month at 500,000 searches. Typesense self-hosted is the server, though its in-memory index sets a RAM floor: budget for an instance that holds the whole index.

    effort: One to two weeks.all Algolia alternatives →
  • Amazon S3MinIO

    S3 storage is around $0.023/GB/month but egress at roughly $0.09/GB is what actually hurts. MinIO on your own hardware or a VPS has no egress charge at all, which for a media-heavy application is usually the entire saving.

    effort: One to two weeks including the data copy.all Amazon S3 alternatives →
  • Astronomer (Astro)Apache Airflow

    Astro deployments start near $0.35/hr with workers from $0.13/hr, and real multi-environment setups are quoted at $1,500–$5,000 a month. Self-hosted Airflow on Kubernetes costs the cluster — often a few hundred.

    effort: Two to six weeks depending on how well you already run Kubernetes.all Astronomer (Astro) alternatives →
  • Atlassian StatuspageGatus

    Atlassian Statuspage starts around $29/month and business tiers reach several hundred. Gatus is Apache-2.0 on a small VPS.

  • Auth0Keycloak

    Auth0 bills per monthly active user and the B2C tiers escalate quickly past the free 25,000 MAU. A 100,000-MAU app commonly pays four figures monthly. Keycloak is free; you pay for the servers and the operating time.

    effort: Three to eight weeks. Identity migrations deserve caution.all Auth0 alternatives →
  • AWS GPU InstancesRunPod

    H100 capacity is about $12.29 per GPU-hour on-demand at AWS against roughly $1.99 at RunPod — a five-to-six-fold difference for identical silicon. A single H100 running 8 hours a day drops from about $2,950 a month to $480.

    effort: Days for a containerised workload. Weeks if it is wired into SageMaker.all AWS GPU Instances alternatives →
  • DatadogGrafana + Prometheus

    Datadog bills roughly $15–23 per host per month for infrastructure plus $0.10 per ingested GB of logs, and real bills land between $200 and $2,400 a month. Grafana plus Prometheus on a $20 VPS costs the VPS. A 20-host estate typically drops from about $600/month to under $50.

    effort: Two to four weeks part-time for a team that already runs Kubernetes. Longer without.all Datadog alternatives →
  • DatadogSigNoz

    SigNoz self-hosted is free and stores in ClickHouse, which compresses observability data 10–20× better than Elasticsearch. Teams report 80% storage cost reductions against SaaS. Against a $600/month Datadog bill, expect $40–80 in infrastructure.

    effort: One to two weeks if already on OpenTelemetry. Three to four if migrating off the Datadog agent first.all Datadog alternatives →
  • Docker DesktopPodman + Podman Desktop

    Docker Desktop requires a paid subscription for businesses over 250 employees or $10m revenue — Pro is $9/user/month, Team $15. Podman Desktop is Apache-2.0 and free at any size.

    effort: Days for individual developers. Weeks to standardise a team.all Docker Desktop alternatives →
  • FirebaseSupabase

    Firebase Blaze bills per operation — about $0.06 per 100,000 reads — so a read-heavy app can produce a surprising invoice. Supabase Pro is a flat $25/month, or free self-hosted. The saving is unpredictability as much as amount.

    effort: Three weeks to two months depending on Cloud Function count.all Firebase alternatives →
  • FirebasePocketBase

    PocketBase is one Go binary with embedded SQLite — it runs on a $5 VPS. Against a $50–200/month Firebase bill for a small app, that is the whole saving.

    effort: Under two weeks for a small app.all Firebase alternatives →
  • GitHub ActionsWoodpecker CI

    Standard Linux runners are about $0.006/min and macOS about $0.062/min — roughly ten times more. A team burning 30,000 macOS minutes a month pays about $1,860; the same work on a self-hosted Mac mini is the hardware once.

    effort: A week for a simple pipeline. Longer where marketplace actions are load-bearing.all GitHub Actions alternatives →
  • HashiCorp VaultOpenBao

    Neutral on licence cost if you already self-host Vault Community. Against HCP Vault — from about $450/month for a dev cluster and roughly $1,150 for production Essentials — OpenBao is the infrastructure only.

    effort: Days to two weeks depending on how much of Vault's surface you use.all HashiCorp Vault alternatives →
  • HashiCorp VaultInfisical

    Infisical's MIT core is free to self-host. Against HCP Vault at $450–$1,150 a month the saving is the whole subscription; against self-hosted Vault the saving is operational time rather than money.

    effort: Under a week for a key-value use case.all HashiCorp Vault alternatives →
  • HerokuCoolify

    Heroku's cheapest production dyno plus a Postgres add-on starts near $50/month per app and rises steeply. Coolify on a $6–14 Hetzner or DigitalOcean VPS hosts a dozen containerised apps for the price of the box.

    effort: A weekend for a straightforward app; longer with many add-ons.all Heroku alternatives →
  • LaunchDarklyUnleash

    LaunchDarkly bills roughly $10 per service connection plus about $8.33 per 1,000 client-side MAU. A consumer app with 200,000 MAU pays roughly $1,700/month. Unleash self-hosted is a server and a database.

    effort: One to three weeks, most of it flag audit rather than engineering.all LaunchDarkly alternatives →
  • LaunchDarklyFlagsmith

    Same LaunchDarkly baseline. Flagsmith self-hosted is free under BSD-3-Clause with no per-MAU meter at all.

    effort: One to two weeks.all LaunchDarkly alternatives →
  • ModalvLLM

    Modal's H100 is about $3.95/hr against roughly $1.99 renting directly and running vLLM yourself — you stop paying the orchestration premium. For an endpoint busy 12 hours a day that is roughly $1,400/month against $700.

    effort: Under a week for a single model behind a single endpoint.all Modal alternatives →
  • MongoDB AtlasPostgreSQL

    A dedicated M40-class Atlas cluster runs roughly $3,200/month; the equivalent three-node self-hosted setup on AWS is about $2,156, saving around $12,500 a year against roughly $5,000 of first-year setup. Moving to managed Postgres instead is usually cheaper again.

    effort: One to three months. The query rewrite dominates.all MongoDB Atlas alternatives →
  • MongoDB AtlasFerretDB

    FerretDB is Apache-2.0 and free; you pay for the Postgres underneath. Against a $3,200/month Atlas cluster, a well-sized managed Postgres runs several hundred. The saving is smaller than going fully relational but the work is a fraction.

    effort: One to two weeks for a straightforward document store.all MongoDB Atlas alternatives →
  • New RelicSigNoz

    New Relic bills per user plus $0.35 per GB ingested beyond 100GB free. A team of eight with moderate ingest lands near $400/month. SigNoz self-hosted is infrastructure only — typically $50–100/month on a box sized for ClickHouse.

    effort: One to three weeks depending on instrumentation.all New Relic alternatives →
  • PingdomUptime Kuma

    Pingdom starts around $10/month for 10 checks and rises with checks and users. Uptime Kuma runs on a $5 VPS with unlimited checks.

    effort: An afternoon.all Pingdom alternatives →
  • Red Hat OpenShiftk3s

    OpenShift is licensed per core or socket and enterprise agreements run to five and six figures annually. K3s is Apache-2.0 and free; you pay for nodes.

    effort: One to three months for a real estate.all Red Hat OpenShift alternatives →
  • RedisValkey

    Neutral if you self-host — Valkey is a drop-in fork. Against Redis Cloud or Redis Enterprise the saving is the whole licence, and against managed Redis it is whatever your provider charges for the Valkey equivalent, usually less.

    effort: An afternoon for plain Redis. A project if you use Stack modules.all Redis alternatives →
  • ReplicatevLLM

    From per-second GPU billing with a platform margin, to the raw cost of a GPU you rent or own. At steady utilisation this is commonly a two-thirds saving; at spiky or low volume it is a loss, because scale-to-zero genuinely costs nothing between requests.

    effort: a day to prototype, a week to run in production with confidenceall Replicate alternatives →
  • SegmentJitsu

    Segment bills per monthly tracked user and mid-market plans reach four figures. Jitsu self-hosted is infrastructure — typically $50–150/month for the same event volume.

    effort: Two to four weeks.all Segment alternatives →
  • SendGridPostal

    SendGrid's Essentials starts near $20/month for 50,000 emails and scales up quickly. Postal self-hosted is a VPS with good IP reputation — around $20/month regardless of volume.

    effort: Two to four weeks, most of it IP warm-up rather than engineering.all SendGrid alternatives →
  • SentryGlitchTip

    Sentry bills by event volume and teams commonly land at $80–$500/month. GlitchTip self-hosted is a small VPS — $10–20 — and speaks the same protocol.

    effort: An afternoon.all Sentry alternatives →
  • SnowflakeClickHouse

    Snowflake bills per credit for compute plus storage, and idle warehouses that auto-resume produce bills nobody predicted. ClickHouse self-hosted is the machine — often an order of magnitude less for analytics-shaped workloads.

    effort: One to three months for a real warehouse.all Snowflake alternatives →
  • SplunkOpenSearch

    Splunk prices by daily ingest volume and enterprise agreements routinely reach six figures. OpenSearch is Apache-2.0 and free; you pay for the cluster. A 100GB/day estate that costs tens of thousands annually on Splunk runs on hardware in the low thousands.

    effort: Three to six months for a real enterprise estate. This is the heaviest migration on this site.all Splunk alternatives →
  • AWS TextractDocling

    From $15–$65 per 1,000 pages for table and form analysis to zero per page. A 100,000-page corpus ingest goes from a four-figure invoice to a weekend of compute — and, more importantly, re-running it after a chunking change costs nothing instead of the same amount again.

    effort: two to three days including the re-ingestall AWS Textract alternatives →
  • Azure AI Content SafetyNVIDIA NeMo Guardrails

    From per-1,000-records pricing on every message in both directions, to free plus whatever compute the rails use. High-traffic conversational products are where this shows, because you moderate input and output on every turn.

    effort: a week including the shadow run, and do not skip the shadow runall Azure AI Content Safety alternatives →
  • BraintrustPromptfoo

    From roughly $249 a month per small team plus trace overage, to free with unlimited seats. The seat count matters more than the headline: everyone who edits a prompt needs to run evals, so the correct licence count is your whole engineering team.

    effort: a day for the first ten cases, a week to port a mature suiteall Braintrust alternatives →
  • Deepgramfaster-whisper

    From roughly $0.0043 a minute to nothing per minute. A product transcribing 1,000 hours a month goes from about $260 to the cost of the electricity, and a one-off archive ingest stops being a four-figure invoice entirely.

    effort: a weekend for batch, longer if you need streamingall Deepgram alternatives →
  • ElevenLabsPiper

    ElevenLabs bills per character and production usage reaches hundreds monthly. Piper runs on CPU — a Raspberry Pi manages real-time synthesis.

  • LangChainDirect SDKs (no framework)

    Neutral on spend — same model calls either way. What you remove is a dependency that has broken compatibility across major versions more than once, and roughly 100MB of transitive packages.

    effort: Days for a straightforward chain. Longer if you built on agents.all LangChain alternatives →
  • LangSmithLangfuse

    LangSmith bills per trace and teams with real volume reach hundreds monthly. Langfuse self-hosted is Postgres and a container.

    effort: Under a week for tracing. Longer if you have substantial evaluation sets.all LangSmith alternatives →
  • OpenAI AgentKitLangGraph

    Roughly neutral on model spend — you still pay per token. What you avoid is a product being withdrawn: OpenAI announced on 3 June 2026 that Agent Builder is winding down and leaves the platform on 30 November 2026.

    effort: One to four weeks depending on workflow count.all OpenAI AgentKit alternatives →
  • OpenAI API (ChatGPT)Ollama

    GPT-class API usage at moderate volume runs $50–500/month. Ollama on hardware you own is electricity. On a rented GPU it is roughly $200/month for a card that serves a small team continuously.

    effort: An afternoon to try. Weeks to tune properly.all OpenAI API (ChatGPT) alternatives →
  • OpenAI API (ChatGPT)vLLM

    At scale the comparison flips: an H100 at roughly $2/hr serving continuously costs about $1,400/month and handles throughput that would cost far more per token via API. Below heavy sustained use, the API is cheaper.

    effort: A week to serve. Ongoing to operate.all OpenAI API (ChatGPT) alternatives →
  • OpenAI Embeddings APINomic Embed

    OpenAI charges about $0.02 per million tokens for text-embedding-3-small and $0.13 for the large model. At a million documents re-embedded monthly that is real money; nomic-embed runs on your own hardware or a CPU box for nothing.

    effort: Days for the swap. The re-embedding run itself scales with corpus size.all OpenAI Embeddings API alternatives →
  • OpenAI Fine-TuningAxolotl

    From per-token training plus a permanent per-call inference premium, to a few dollars of rented GPU time and then nothing. The durable saving is the inference premium: a fine-tuned OpenAI model costs more per call than its base model, forever.

    effort: a few days including the learning curve, hours once you have done oneall OpenAI Fine-Tuning alternatives →
  • PineconeQdrant

    Pinecone's serverless pricing scales with storage and reads, and production indexes commonly run $70–500/month. Qdrant self-hosted is the server — a single node handles millions of vectors on modest hardware.

    effort: One to two weeks.all Pinecone alternatives →
  • PortkeyLiteLLM

    Portkey's paid tiers scale with request volume. LiteLLM self-hosted is a proxy container — effectively free.

  • VectaraLlamaIndex

    From about $50 a month plus query-based usage, to free plus whatever you spend on embeddings and a vector store. The structural change matters more than the number: query-based pricing scales with your product's success, so the better it does the more retrieval costs.

    effort: one to two weeks done properly, and shortcuts here are false economyall Vectara alternatives →
  • Weights & BiasesMLflow

    From about $50 per user per month plus storage and tracked-hour overage, to free. A six-person team goes from roughly $3,600 a year plus storage to the cost of a small server and an object bucket — call it $20 a month.

    effort: a day of code, plus however long the artefact download takesall Weights & Biases alternatives →

Common questions

What is in a Macrostack migration guide?
Five things, in the order you need them: what the move does to your bill, what data carries over cleanly, what you will have to rebuild by hand, the actual step sequence, and the one thing that catches people out. Each guide ends with a written 'when to stay' — the conditions under which the tool you are already paying for is the right answer.
Are these guides written from real migrations?
They are written from the documented export and import paths of both products, checked against each vendor's own migration and API documentation. Where a path does not exist and the answer is genuinely 'you rebuild this by hand', the guide says so rather than inventing a tool that would do it.
Do you only recommend leaving paid tools?
No, and that would make this a worse resource. Every guide carries a 'stay if' section, because for a lot of teams the incumbent is the correct choice — the managed backup, the compliance paperwork and the on-call rota are worth real money. Macrostack is a decision engine, not an exit engine.
How many migration guides are there?
65, spread across 3 layers of the stack. They are weighted toward infrastructure and models, because that is where the bills are large enough for a move to be worth a weekend.
The Macrostack brief

New swaps, worth your inbox.

A short, occasional email when we add a high-intent alternative or ship a new head-to-head. No spam, no selling your address — unsubscribe in one click.