★ 41,098TypeScriptApache-2.0updated Oct 2026
best for teams that need the widest connector coverage without picking a niche tool
Connector breadth is still the strongest reason to pick Appsmith over NocoDB or Directus — 25+ databases plus any REST or GraphQL API — though that same breadth, paired with a widget tree that re-evaluates in full on every state change, is why canvases with real complexity start to feel sluggish in a way the narrower tools don't.
Self-hosting is unusually solid for an open-source tier: Docker Compose and a maintained Helm chart both work out of the box, and the CI pipeline covers Cypress, Playwright, and visual regression tests that most competitors skip. What it doesn't cover for free is RBAC, audit logs, or SSO — all three sit behind the commercial tier, so a team of five outgrows the free version faster than ToolJet or Budibase users do.
Our full take →GitHub ↗
★ 41,113JavaScriptAGPL-3.0updated Oct 2026
best for teams with multiple developers who need to edit the same app together
ToolJet's multiplayer editing — several developers in the same app at once — is a feature Appsmith and Budibase both skip, and the 80+ connector list covers roughly the same ground Appsmith does. Getting there means accepting AGPL, which is a harder license to clear with legal than Appsmith's commercial-tier gating or Budibase's MPL terms.
Multi-environment support, Git sync, and row-level access are all paid-only, and the README now opens with an AI app-generation pitch that makes the free product harder to find under the marketing.
Our full take →GitHub ↗
★ 65,312TypeScriptNOASSERTIONupdated Oct 2026
best for teams that want an Airtable-style spreadsheet over a database they already run
NocoDB's pitch is narrower than Appsmith's or ToolJet's — point it at an existing Postgres, MySQL, or SQLite table and it gives you a spreadsheet UI with grid, kanban, calendar, and form views, plus an auto-generated REST API with Swagger docs, no connector-writing required. Field-level role-based access goes further than most of the self-hosted alternatives here.
Two things will trip up a migration: formula support lags Airtable's, so a nested formula that worked there may not translate, and performance tails off somewhere past 50k rows per table, where the UI starts to stutter. The license is Sustainable Use rather than MIT, and it puts real limits on commercial use that aren't obvious from the README.
Our full take →GitHub ↗
★ 38,641TypeScriptNOASSERTIONupdated Oct 2026
best for teams with an existing database who want an API and admin UI without a migration
The permission model is powerful but it takes most teams a week of role, policy, and access layering before it stops feeling foreign, and the license moved to MSCL, a source-available license rather than open source, gating 'advanced' features behind a revenue or headcount line Directus draws itself. None of that changes what it's good at: wrapping whatever database you already run — Postgres, MySQL, MSSQL — in a REST/GraphQL API and a Vue admin UI without migrating the schema first.
Auth is production-grade from day one, with OAuth2, OpenID Connect, SAML, and LDAP all built in rather than bolted on later, and real-time WebSocket subscriptions ship free, something a few of these charge for. GraphQL mutations on deeply nested data are still less reliable than the REST API, so anything complex is safer built against REST first.
Our full take →GitHub ↗
★ 21,910TypeScriptNOASSERTIONupdated Oct 2026
best for teams that need an Airtable alternative at a scale NocoDB can't hold
Teable's answer to NocoDB is scale — it's demoed at a million rows, well past where NocoDB starts to stutter, because the data lives in real Postgres tables rather than a JSON blob column. The plugin SDK also gives developers an extension point Airtable itself doesn't offer self-hosters, and the dual Postgres/SQLite backend is cleaner engineering than most of the field.
Automation and AI, the features teams reach for once they outgrow a spreadsheet, sit behind an AGPL Enterprise Edition, and Redis is a hard dependency for the cache and queue layer — skip it and jobs fall back to a local queue that silently drops work on restart. Docker Compose upgrades have no documented migration path when the Prisma schema underneath changes.
Our full take →GitHub ↗
★ 28,336TypeScriptNOASSERTIONupdated Oct 2026
best for teams that need a non-copyleft license for apps they'll ship commercially
Budibase settles the license question more simply than ToolJet or Teable — the component and client libraries are MPL, so apps built on it don't inherit the AGPL copyleft exposure those two carry. Self-hosting covers Docker, Compose, Kubernetes, and a single-image option, and the connector list (Postgres, MySQL, REST, S3, Google Sheets) matches Appsmith's.
The database underneath is CouchDB, an unusual foundation next to the Postgres and MySQL most of this list runs on, and a bundled custom SQLite index binary suggests the team has been working around CouchDB's limits for a while. The recent pivot toward AI agents that act on employee requests sits on top of an existing low-code architecture, and how much control you have over what those agents can do is thin.
Our full take →GitHub ↗
★ 24,652TypeScriptNOASSERTIONupdated Oct 2026
best for teams planning to extend the platform with their own plugins
The open-source line isn't documented anywhere — you find it by noticing build-pro-image targets buried in the CI config, not a pricing page — and the monorepo is large enough that plugin development means reading undocumented internal APIs. The 'AI employees' framing oversells it too: what's running under the hood is configuring which tools an LLM can call, not an agent making business decisions on its own.
What's underneath holds up, though: the plugin architecture is the most consistent of the group, every feature including core ones is a plugin, so new functionality follows one pattern instead of forking into a monolith. Data stays in a relational database you can query directly with SQL, and the field-level permission system applies the same rules to human users and AI agents, closing a gap Budibase's agent pivot leaves open.
Our full take →GitHub ↗
★ 18,237RustNOASSERTIONupdated Oct 2026
best for teams turning existing scripts into webhooks, schedules, and UIs
Windmill starts from a different place than the rest of this list: write a typed Python, TypeScript, Go, or Bash function and it generates the UI, webhook, or schedule from the signature, instead of you assembling widgets on a canvas. Postgres as the job queue instead of Redis or Kafka is the right call at this scale, and nsjail with PID namespace isolation is a real security boundary for a platform running arbitrary code against production credentials.
The license is murkier than the badge suggests — it reads AGPLv3, but the Community Edition Docker image ships proprietary code alongside the open parts, so anything redistributable needs a commercial license regardless of what the badge implies. The '13x faster than Airflow' number compares against a tool built for a different job, data-pipeline DAGs rather than low-latency dispatch, and there's no stated point at which Postgres-as-queue runs out of headroom.
Our full take →GitHub ↗
// questions people ask
Before you choose
Which of these can I run without Kubernetes?
All eight have a plain Docker or Docker Compose path that works for a single team — Kubernetes and Helm charts exist for scale, not as a requirement to get started. Appsmith, ToolJet, Budibase, and Windmill all document Compose as the default setup, not an afterthought.
Can non-developers build apps on these without engineering involved?
NocoDB and Teable come closest, since both are spreadsheet-first interfaces over a real database that a non-technical person can drive alone. The rest — Appsmith, ToolJet, Budibase, Directus, NocoBase, Windmill — assume someone with engineering skills is doing the building, even if end users just consume the result.
Which license should worry me if I'm building something I'll eventually sell or redistribute?
AGPL on ToolJet, Teable's app layer, and Windmill's core, plus the proprietary code bundled into Windmill's Community Edition image, all need a real legal read before you build a redistributable product on them. Budibase's MPL client libraries and Appsmith's core are simpler starting points for that; Directus and NocoDB use source-available terms that restrict commercial use directly rather than through copyleft.
What typically gets paywalled first once a team outgrows the free tier?
SSO, audit logs, and row-level or field-level access control get gated first on Appsmith, ToolJet, Teable, and Budibase. Directus and NocoBase instead gate by revenue or deployment scale rather than by feature, which is harder to see coming.
If you already run Postgres and want to avoid license fine print, Appsmith or NocoDB are the safer starting points. If the job is turning scripts you already have into webhooks and internal UIs, Windmill is built for exactly that.