mrkeyoor.com_
Sat 17 Oct 13:34 UTC
Self-Hostedevaluationupdated 11 Oct 2026

budibase review

Budibase lets you build internal web apps and automate work around their data, such as recording requests and routing approvals. It also has AI agents that can act through those workflows and connected business systems.

-3stars / 7d
Verdict

Our Budibase run installed 2,549 packages and used 2,036 MB, then passed its build but ended the test command with exit code 130 despite passing package suites. Trial it when an internal app needs approval workflows and connected business actions enough to justify owning the platform. Skip it when a shared table covers the job, or when its licensing and reported keyboard-control gap conflict with your requirements.

We ran it

Lab card: what happened when we ran budibaseScreenshot of budibase (budibase.com)
Install✓ · 189s2549 packages · 2036 MB
Build✓ · 109s
Tests✗ · 15s164 passed · 0 failed of 164 (jest)
Repo4510 files~577,550 lines of source · 143.2 MB · 20 CI workflows

Answers from our run

Does budibase build from source?

Dependencies installed in 189 seconds (2549 packages), and the build succeeded in 109 seconds. We cloned commit ee2f184 into a clean Debian container with 3 CPUs and no project-specific setup.

Do budibase's tests pass?

No. The project's own test command finished in 15 seconds and reported 164 passed, 0 failed (jest).

Who should not use budibase?

Teams requiring permissive licensing across the platform: the README separates GPL v3 platform code, MPL client libraries, and Business Source License paid features.

What are the alternatives to budibase?

Appsmith, NocoDB. Our Budibase run installed 2,549 packages and used 2,036 MB, then passed its build but ended the test command with exit code 130 despite passing package suites.

Setup3/5Install and build passed; services and secrets need deliberate setup
Docs4/5Separate hosting and contributor guides explain services and storage
Community5/528,334 stars, an October 11 push, and current issue activity
Maturity3/5Active v3.48.0 release; incomplete suite and reported keyboard gap

Discussed on

  1. hnShow HN: Budibase – An open-source low code platform280 points
  2. hnMicro SaaS Ideas You'll Want to Steal36 points
  3. hnShow HN: Budibase – open-source Retool alternative – launch their Public API30 points
  4. hnShow HN: Budibase 2.0 – build internal tools and forms faster (OSS)26 points
  5. hnBudibase raises $7M for its open source low-code platform14 points

Who it’s for

Operations teams whose employee requests need forms, stored records, and approval workflows in the same system.
Engineers who want to connect internal apps to existing databases and business services.
Teams that can own persistent storage and service configuration, or choose Budibase Cloud.

Who it’s NOT for

Teams requiring permissive licensing across the platform: the README separates GPL v3 platform code, MPL client libraries, and Business Source License paid features.
Teams requiring confirmed keyboard access before adoption: open issue 19847 reports mouse-only cells and pagination controls in shared BBUI components.
Developers expecting to run the platform with Node alone: the contributor guide requires Docker and Compose for backend services.
Teams whose fork depends on unrestricted upstream submissions: external PRs need an assigned, triaged issue and a CLA, and contributors may have only one open PR.

Setup reality

Our sandbox installed Budibase commit ee2f184 in 189 seconds, adding 2,549 Yarn packages and occupying 2,036 MB. The build passed in 109 seconds. Tests failed after 15 seconds with exit code 130. The recorded summary has 164 passed and 0 failed; the tail separately shows 164 shared-core and 395 string-templates tests passing. That is a partial result, not a completed suite.

Running the platform needs persistent storage and configured service secrets. The current Compose route includes app and worker services, a proxy, CouchDB, MinIO, Redis, LiteLLM, and PostgreSQL. External databases and business integrations need their own access settings, and email needs SMTP configuration.

Contributor setup calls for Node 22, Python 3, Docker, and Compose; Windows users must run Yarn commands in Bash. The single-image route still needs a persistent /data mount and production secret configuration. Our successful source build does not verify a running deployment or production sizing.

Budibase connects request forms to workflow actions

Budibase is worth considering when an employee's request needs to create a record, reach an approver, and update an internal app. The README puts agents alongside apps and automations, with database and business-system connections beneath them. That makes an approval process a better starting point than an isolated chatbot. The README describes those actions.

Taking ownership of that platform has a visible cost. Our checkout contained 4,510 files and roughly 577,550 source lines, occupying 143.2 MB before dependencies. The contributor guide separates the Svelte builder, browser client, Koa server, and worker. If you only need colleagues to edit a table, that's a lot of application machinery to adopt. Budibase makes more sense when the same data needs screens and repeatable actions.

What happened when we ran it

Our run installed Budibase commit ee2f184 in 189 seconds, pulling 2,549 Yarn packages and occupying 2,036 MB on disk. The build succeeded in 109 seconds. We cloned it on October 2, 2026, into a fresh, unprivileged Debian sandbox with 3 CPUs, 8 GB of RAM, the lab-node:22 image, and no secrets. Installation and compilation completed under those conditions.

The test command failed after 15 seconds with exit code 130. The recorded Jest summary shows 164 passed and 0 failed out of 164. The log tail gives more context: shared-core finished 17 suites with 164 passing tests, and string-templates finished 10 suites with 395 passing tests. Those are separate package results. They do not establish that the entire workspace test command completed successfully. The final output names the backend-core test task and repeats Yarn's exit-code error, without explaining why the command stopped. We can't assign that result to missing system packages, memory pressure, or a code defect. The finding is narrower: successful installation and build, passing package suites in the log, and an unsuccessful overall test step. The 15-second result is insufficient evidence for approving the full suite.

Docker packages the platform, while storage remains your job

The single-image Docker guide starts with a persistent /data mount. Its production settings cover session signing, internal API access, Redis, CouchDB, and MinIO credentials. Configure those deliberately and supply credentials for the external systems your apps will touch. SMTP also needs configuration if you use email. A convenient launch command doesn't answer how you'll restore the app's data after a host failure.

The current Compose file separates app and worker services from the proxy, CouchDB, MinIO, Redis, LiteLLM, and its PostgreSQL database. Several stores have persistent volumes. Our 2,036 MB installed footprint describes the source environment, not this running cluster. Likewise, the sandbox's 8 GB of RAM is a measurement condition, not a production sizing recommendation. Choose the deployment route before budgeting the server.

GPL, MPL, and paid-feature licensing need separate decisions

The README identifies GPL v3 for Budibase, MPL for client and component libraries, and the Business Source License for paid features. It says apps you build can use your chosen license. That's useful, but a team requiring permissive licensing throughout the platform should walk away. A team needing a paid capability should confirm its terms before building a workflow around it. Self-hosting doesn't settle that question.

Maintaining a fork is another commitment. The contributor guide requires a CLA and, for external pull requests, an open, triaged issue assigned to the contributor. It limits contributors to one open pull request each. Against the roughly 577,550 source lines we measured, those rules matter if your adoption depends on upstream accepting local changes. Discuss the required change before promising it to your users.

Shared table controls have a reported keyboard-access gap

Issue 19847, opened October 11, 2026, reports that BBUI clickable cells and pagination arrows use non-focusable elements without equivalent keyboard handling. The report concerns shared components and says the proposed Actions interface inherits the limitation. It doesn't establish that every Budibase app is inaccessible. It does give accessibility-sensitive teams a specific interaction to check before choosing the platform.

The 164 passing tests in our recorded summary cannot settle that keyboard question. They belong to shared-core, while the issue concerns BBUI controls. If an employee needs to navigate a table without a mouse, evaluate that actual screen and its pagination in the deployment you intend to use. Treat the issue as a reported limitation requiring verification, rather than evidence that an entire product fails an accessibility standard.

October activity shows maintenance beyond the release tag

GitHub reported 28,334 stars and 270 open issues and PRs, with the last push on October 11, 2026. Release v3.48.0, published October 5, includes approval-flow work, app-navigation changes, and fixes around batch deletion and publishing locks. Together with newly opened component issues, that supports an active-maintenance judgment.

Our scan found 20 CI workflow files and monorepo workspaces, while reporting no Dockerfile or tests directory. That last signal plainly doesn't mean there are no tests: the supplied log contains completed Jest suites. Nor does the scan erase the published Docker hosting route. These are repository-discovery signals with limits. For a release decision, use the actual command outcomes and inspect the affected workflow rather than treating file counts as certification.

A table-only job deserves a narrower candidate

Appsmith is worth comparing when an admin panel or dashboard over existing databases and APIs is the main requirement. NocoDB is the closer candidate when people primarily need spreadsheet-style records, forms, and views. We haven't measured those alternatives here. Choose between them by the work employees must complete, then evaluate the relevant deployment.

I'd trial Budibase with an approval workflow that already causes manual handoffs. Budget for the 2,036 MB source environment if you'll modify the platform, and assign ownership of its persistent stores before employees depend on it. Require a completed test run for your chosen revision. The pilot should end with someone submitting a request and the right person receiving it, with the resulting record available to inspect.

Alternatives

ProjectWhat it isPick it when
Appsmith gh↗A visual builder for admin panels, dashboards, and internal apps connected to databases and APIs.pick this instead when a database-backed admin interface is the main job and you want to compare its app-building approach.
NocoDB gh↗A self-hostable database interface with spreadsheet-style editing, forms, and multiple views.pick this instead when colleagues mainly need to manage shared records rather than build an operations platform.

Sources

  1. Budibase README and licensing split
  2. Contributor setup and submission rules
  3. Single-image Docker hosting and production settings
  4. Current Docker Compose services
  5. Budibase v3.48.0 release
  6. BBUI keyboard accessibility report, issue 19847
  7. Appsmith repository
  8. NocoDB repository

More self-hosted reviews

playport · edge-smart-doh · Comma · ProsperoEden · zulu · node_exporter · the whole board →