This is a source and documentation review of NocoBase, based on repository, release, and issue data collected on 2026-08-15; no runtime testing was performed for this article. In brief: NocoBase is an open-source, TypeScript-based AI + no-code platform for building business systems such as internal tools, CRMs, and ERP-style applications, and it is actively maintained, with stable release v2.1.40 published on 2026-08-14. It is worth serious evaluation if you want a self-hosted, data-model-driven platform and can own deployment, upgrades, and license review. Before production, verify the license terms (GitHub reports the license as Other with SPDX NOASSERTION), test the specific blocks and workflows you depend on, and plan an upgrade path, because some UI bugs have remained open for months even while releases ship frequently.

Key facts at a glance

What NocoBase is

According to the repository description, NocoBase is an open-source AI + no-code platform for building business systems fast, with AI working on top of production-proven infrastructure and a WYSIWYG no-code interface. The maintainers describe built-in data models, permissions, workflows, and audit logs; relational data decoupled from the UI; and a plugin-based microkernel architecture. The README also lists support for coding agents such as Claude Code, Cursor, Codex, OpenCode, and TRAE, and describes AI employees with role-based permissions and audit logs. Those are maintainer claims from the README, not results verified in this review.

What changed in v2.1.40

v2.1.40 is a bug-fix and improvement release published on 2026-08-14 by the nocobase automation bot. The main improvement is configurable default sorting for tree filter blocks. Bug fixes span client v2 filter forms, JSON calculation workflows, exports, kanban, file manager, import records Pro, and approval workflows, including:

• Filter forms refreshing automatically in UI edit mode, tree-structured relation fields defaulting to Cascader, mobile single-select relation fields not clearing, China region fields in filter forms, and stray JS items in form linkage rules.
• JSON Query workflow jobs losing structured parser errors and displaying [object Object].
• Export V2 page crashes when sorting an edited relation field, and export errors after filtering by multiple relation fields.
• Kanban card content disappearing after reopening a detail drawer, and video previews not updating in the file manager.
• Import records Pro not associating asynchronously imported records with the current space, plus clearer results separating newly created from updated records.
• Approval workflow subtable column widths and data scope filtering in Client V2 approval blocks.

This is a maintenance release, not a feature release. The signal for operators is that the v2.1 line is receiving steady fixes across filters, exports, workflows, and approval blocks.

What the release cadence tells you

In the evidence window, NocoBase published v2.1.39 on 2026-08-11 and v2.1.40 on 2026-08-14, while also publishing v2.2.0-beta.17 on 2026-08-10 and v3.0.0-alpha.9 on 2026-08-11. Active development spans a stable 2.1.x line, a 2.2 beta line, and a 3.0 alpha line. If you adopt NocoBase today, plan for upgrade testing across lines rather than assuming a single long-term support channel.

Who should consider NocoBase

NocoBase is a reasonable candidate if your team needs a self-hosted platform for internal tools, CRM, ERP, admin dashboards, or workflow-heavy applications, and you have someone who can operate a server and database. The repository’s topics match that profile: self-hosted, workflows, ai-agent, crud, erp, crm, and internal-tools. The data-model-first, UI-decoupled design and plugin architecture suit teams that expect their data model to outlive any single UI iteration.

Conditions matter. If you want a zero-maintenance SaaS subscription, or if your organization requires a standard open-source license such as MIT, Apache, or GPL, NocoBase needs a licensing review first, because GitHub reports the license as Other with SPDX NOASSERTION. If your core workflow depends on a specific block such as calendar, approvals, exports, or relation-driven forms, check the open issues and recent release notes for that area before committing, since those have been active bug areas.

Maintenance signals

Maintenance looks healthy on release activity but uneven on bug closure. On the positive side, the repository was actively pushed on 2026-08-14, the stable release line ships frequently, and the snapshot shows 23,653 stars and 2,824 forks, a sign of broad adoption.

On the cautionary side, 292 issues were open at collection time, and two examples show that bugs can linger. Issue #7625, where the calendar view does not switch to the correct days, was opened on 2025-10-20 and remained open on 2026-08-15. Issue #9050, a PostgreSQL error when searching relation tables with formula fields, was opened on 2026-04-07 and remained open. Both had assigned maintainers but had not closed by the snapshot date. Frequent releases do not guarantee every reported bug is fixed quickly, so validate your specific feature set.

Deployment concerns to check before production

• License and commercial terms. GitHub reports the license as Other with SPDX NOASSERTION. Before production, have the license text and any commercial edition terms reviewed, and confirm what the community versus paid feature split means for you; release notes show Pro plugins such as Import records Pro and Export Pro.
• Self-hosting and operations. The README markets self-hosting and states that data stays in your own database with no platform lock-in; that is a maintainer claim. You still own backups, database administration, security patching, and monitoring.
• Upgrades. With 2.1.x stable, 2.2 beta, and 3.0 alpha lines active, pin a version, read release notes, and test migrations in staging. Prerelease notes include migration-related fixes such as many-to-many through-table data after restarts, so upgrade testing is not optional.
• AI features add dependencies. The README describes AI employees, MCP and HTTP APIs, and coding-agent support; release notes reference LLM service setup and vector databases for knowledge bases. Plan for external AI services and their cost, latency, and data-handling implications. These capabilities are maintainer-described and were not verified at runtime in this review.
• Database and data source specifics. Recent fixes target PostgreSQL, MSSQL, and external MySQL sources, plus Sequelize-related behavior. Match your database choice to documented support and test relation-field search and exports, which have been active bug areas.

Limitations and what is unknown

This review is a source and documentation review. There is no local installation, benchmark, or production deployment evidence behind it, so performance under load, migration quality for existing legacy data, and the real-world behavior of AI features are unknown here. The README’s claims about production-proven infrastructure, permissions, and audit logs are maintainer statements, not independently verified results.

Also unknown: the exact license obligations due to the NOASSERTION classification, the release cadence beyond the snapshot, and which features require paid plugins. The project points to documentation, an online demo, and a forum, but those were not evaluated as part of this review.

Evaluation questions and alternatives

Ask these before comparing products:

1. Which business entities, relationships, permissions, and workflows do you actually need, and which already exist in your current system?
2. Who will operate the instance, and is there capacity for upgrades, backups, and incident response?
3. Are the blocks you depend on, such as calendar, approvals, exports, and filter forms, stable in the version you would deploy? Check open issues and recent release notes rather than the feature list.
4. What does your legal review say about the license, and what data will flow to external AI services?
5. What is your exit path: can you export the data, and what custom code would you lose if you switch?

Alternatives depend on the tradeoff you are making. If the priority is minimal operations, a managed no-code or SaaS business platform shifts hosting and upgrades to the vendor. If the priority is code-level control with self-hosting, NocoBase’s plugin architecture and open API surface are more relevant. If you need a packaged CRM or ERP with industry-specific workflows out of the box, purpose-built commercial products may beat a general platform, even though they cost more and constrain customization. The evidence here does not compare NocoBase to specific competitors; treat these as evaluation categories, not benchmark results.

Verdict

NocoBase is worth a pilot for teams that want a self-hosted, data-model-driven platform and can staff its operations. The active release cadence and the breadth of built-in modules are real positives. The blockers are organizational: license review, upgrade planning, and validation of the exact features you depend on, given that some UI bugs have stayed open for months. A sensible first step is a non-critical pilot with a staging environment, a pinned stable version, and a written rollback plan.

Sources

• NocoBase repository
• Release v2.1.40
• NocoBase documentation
• Issue #9050: relation table formula field search error (PostgreSQL)
• Issue #7625: calendar view does not switch days

If you are evaluating NocoBase or modernizing a legacy business system, I can help with license review, deployment and upgrade planning, and custom development around its plugin model. At XP812 I work with founders and engineering teams on this kind of software decision. Reach out with your stack and requirements.