Open Source No Code Platform
A visual builder can produce a working internal tool in an afternoon, but the afternoon is not the part that decides whether you regret it. What matters is where the app runs and who owns the data model once real business records live inside it - most hosted builders meter per seat and keep both the runtime and the tables in their cloud. The open source platforms here let you assemble the same forms, grids, dashboards, and automations over a database you can inspect, then run the result on infrastructure you control. The app stays extendable by an engineer when it outgrows the canvas, and your records never become hostage to a subscription.
Dify
Open-source LLM app development platform for workflows, RAG pipelines, and agent apps
NocoDB
Self-hostable no-code platform that turns any database into a spreadsheet interface
Penpot
Open-source design platform for teams, with self-hosting and open standards like SVG, CSS, HTML, and JSON
Appsmith
Open-source low-code platform for building internal tools, dashboards, and admin panels
ToolJet
Open-source foundation for building internal tools, workflows, and AI agents with a visual app builder
Directus
Wraps any SQL database in instant REST and GraphQL APIs, a no-code admin app, and auth
Budibase
Open-source operations platform for AI agents, workflows, and internal tools
GrapesJS
Free and open source web builder framework for creating HTML templates inside CMS and other apps
Teable
Spreadsheet-like no-code Postgres for building database apps and team workflows
Our picks
Pick by which end you start from - your data, your APIs, or your code.
Spreadsheet over an existing database: NocoDB NocoDB turns a SQL database into grid, gallery, form, kanban, and calendar views, so a team edits records without a custom app. It self-hosts with Docker on SQLite or Postgres, exposes REST APIs and an SDK for programmatic access, and applies role-based permissions with public or password-protected shares.
Internal tools over your APIs: Appsmith Appsmith is aimed at dashboards, admin panels, and service tools built by binding UI components to queries and APIs rather than writing a backend. It connects to more than 25 databases and any API, so it fits data that already lives elsewhere, and it runs self-hosted or on Appsmith Cloud.
Code-first apps and automation: Windmill Windmill suits teams that would rather keep logic in code: it turns scripts in Python, TypeScript, Go, SQL, and more into shareable UIs, webhooks, scheduled jobs, and flows. Parameters are parsed into a generated frontend, and jobs run with sandboxing, which keeps custom actions testable and reviewable.
Instant API over a SQL database: Directus Directus points at a new or existing SQL database and adds REST and GraphQL APIs, authentication, and a no-code admin app with no schema migration. It keeps your real table structure intact across Postgres, MySQL, SQLite, and others, so one database can back a website, an app, and internal tools at once.
What to weigh before you build on the canvas
The runtime model is the choice a visual builder hides most carefully. Some platforms interpret an app at runtime from stored metadata, which makes editing instant but leaves performance and version control harder to reason about; others generate config or code you can review, test, and deploy through a normal pipeline. For a quick admin panel a metadata runtime is fine. For anything customer-facing, ask how forms, permissions, and background jobs behave under load, and whether the same app can be reproduced outside the designer at all.
The data layer deserves a first-class decision, not a default. NocoDB and Directus point at an existing SQL database and keep your real tables intact, so schema ownership and migrations stay where your team already manages them; Teable and Baserow bring their own Postgres-backed store with a spreadsheet feel that non-engineers pick up quickly. Either is defensible, but decide who reviews a schema change when a business user can add a field from the UI, and confirm exports preserve relationships and attachments rather than flattening everything to CSV.
Then look hard at the escape hatches, because useful internal tools always grow custom validation, API calls, and scheduled jobs. Windmill and ToolJet let engineers drop real Python or JavaScript into an app, and Appsmith binds a UI to any API without a hand-built backend. The question is not whether customization exists but whether it is isolated, testable, and deployable without clicking through production. A good fit lets non-engineers build safely while giving engineers a clean path to extend, debug, and back it up.
Related categories
Frequently asked questions
What are open source no-code platforms best at?+
They shine for internal tools: approval flows, CRUD apps, admin panels, operational dashboards, and lightweight portals where the data model changes often and the audience is your own team. NocoDB and Baserow suit spreadsheet-style data work, while Appsmith and ToolJet suit tools wired to existing APIs and databases. They fit poorly for latency-critical products or highly custom user experiences, so test a real workflow before betting a customer-facing app on the canvas.
How much engineering work does a no-code app still take?+
Less front-end scaffolding, not zero engineering. Someone still models the data, reviews permissions, connects an identity provider, sets up backups, and decides how changes move from testing to production. Engineers matter most at the edges - custom code, API integrations, performance tuning, and cleaning up after business users create inconsistent fields. Platforms like Windmill and Lowdefy lean into this by keeping app definitions in code or config that a developer can version and review.
Should the platform bring its own database or connect to mine?+
Both are valid; make the boundary explicit. Directus and NocoDB layer APIs and an editing UI over an existing SQL database and keep your table structure intact, which suits data your other systems already depend on. Teable, Baserow, and Saltcorn bring a managed Postgres store that is simpler to start but becomes a new source of truth. Avoid setups where the visual layer is the only place your business-critical records live.
What is the difference between the spreadsheet-style tools and the app builders?+
Spreadsheet-database hybrids like NocoDB, Teable, Baserow, and APITable start from tables and views - grid, kanban, calendar, form - and are fastest for structured data entry and collaboration. App builders like Appsmith, ToolJet, ILLA Builder, and Lowcoder start from UI components you bind to queries and APIs, which fits dashboards and tools over data that already lives elsewhere. Many teams run one of each rather than forcing a single tool to do both.
Where do visual apps tend to slow down as they grow?+
Watch list views over large tables, cascading formulas, automations that block a user action, and screens that fire repeated API calls. A visual app scales only if the platform exposes control over queries, caching, and background jobs, so test with realistic data volume rather than a sample table. Also check whether you can profile a bottleneck, or whether every performance problem becomes guesswork inside the editor.
What security checks are specific to no-code platforms?+
Focus on the paths where non-engineers can create risk: public forms, file uploads, custom scripts, webhooks, stored secrets, and role changes. Check whether actions run server-side or in the browser, whether permissions apply per app, table, row, and field, and whether there is an audit log. ToolJet's proxy-only data flow and encryption are the kind of control worth confirming - a narrow audit of the login flow says nothing about the workflow engine or plugin system.
How hard is it to move a spreadsheet or Airtable base into one of these?+
Importing rows is the easy part. The work is mapping formulas, attachments, validation, lookup relationships, and record ownership, since spreadsheets often hide business logic in columns nobody documented. Tools that position themselves as Airtable alternatives - Teable, Baserow, APITable, undb - ease the table import, but plan a cleanup pass and a parallel period where users compare old and new before you retire the original.