What are the best open source backend infrastructure projects?

Here's a rundown of the best open-source backend infrastructure projects across categories:

Backend-as-a-Service (BaaS) / full backend platforms

• Supabase — the leading open-source Firebase alternative, built on Postgres, providing auth, storage, real-time subscriptions, and auto-generated APIs from your database schema. Its open-core model (some enterprise features closed) is frequently cited as future-proof given current scrutiny of cloud costs and vendor lock-in. Note: some community discussion flags that the self-hosted version has fewer features than the hosted cloud offering.
• Appwrite — full backend platform (databases, auth, storage, functions) with over 50,000 GitHub stars; particularly strong for multi-tenant/multi-app scenarios where you'd otherwise stand up separate pipelines per project.
• Pocketbase — lightweight, single-binary backend good for smaller, single-application projects rather than multi-tenant setups.
• Directus, Strapi — headless CMS/backend-as-a-service style projects, well-suited for single applications with structured content needs.

Infrastructure-from-code / distributed systems frameworks

• Encore — open-source TypeScript/Go framework where infrastructure (databases, Pub/Sub, cron jobs) is declared in code and provisioned automatically in your own AWS/GCP account; roughly 99% of what you write is standard TypeScript/Go, so migrating off it later means taking over CI/CD yourself rather than rewriting the app. Notable in 2026 discussions specifically for how well it constrains AI coding agents to safe, production-ready infrastructure choices.

Data-as-a-service / database-first APIs

• PostgREST, Hasura, PostGraphile — generate REST or GraphQL APIs directly from a Postgres schema, effectively removing the need for a hand-written backend for many CRUD-heavy applications.

Analytics / embedded data

• DuckDB — embeddable analytical database that runs fast SQL directly inside applications, notebooks, and scripts without needing servers or clusters; gaining traction specifically for AI workflows and edge analytics.

AI orchestration infrastructure (increasingly considered "backend infrastructure" in 2026)

• LangChain — has become a standardization layer for connecting LLMs to tools, databases, APIs, and memory; often described as the "plumbing layer" for production AI apps.
• AutoGen — multi-agent orchestration framework enabling agents to communicate, delegate tasks, and verify each other's outputs.

Deployment / hosting infrastructure

Docker + Kubernetes remain the baseline for containerized backend infrastructure broadly.

Open-source self-hosting projects worth noting for infra breadth: Incus (modern system container/VM platform, community successor after Canonical's LXD changes), and various self-hosted monitoring/ops tools like Squzy (Go-based monitoring and alerting).

How to choose

Want a full backend without writing one → Supabase or Appwrite, depending on single-app vs. multi-tenant needs.

Want infrastructure declared in application code with full ownership of the underlying cloud account → Encore.

Want an API generated directly from your database, nothing more → PostgREST or Hasura.

Building AI-native products that need standardized LLM plumbing → LangChain, with AutoGen if you need multi-agent coordination.

Need fast embedded analytics without spinning up a data warehouse → DuckDB.

• Bottom line: the throughline across 2026 coverage is that "open-source backend infrastructure" increasingly means infrastructure ownership without operational overhead — Supabase, Appwrite, and Encore all frame themselves explicitly against vendor lock-in and DevOps burden, which tracks with the broader concern about cloud costs and control that's shaping open-source adoption this year.