Review snapshot: August 10, 2026. This is a source and documentation review based on the public GitHub repository and its README; no installation, runtime test, or benchmark was performed.

Bottom line: Carbon is an open-source ERP, MES, and QMS platform that the repository describes as built for complex assembly, contract manufacturing, and configure-to-order manufacturing. As of the August 10, 2026 snapshot, it is worth evaluating if your team already works in the TypeScript, React, and Supabase ecosystem and wants an API-first, extensible core. It is not a low-risk package to adopt: GitHub lists the license as Other (NOASSERTION), the README references AGPL with a commercial license for private use, no tagged releases appear in the evidence snapshot, 144 issues were open, and self-hosting requires a sizable service stack. Treat it as an early-stage platform to validate, not a turnkey ERP.

Key facts at a glance

What Carbon actually is

The repository describes Carbon as an open-source ERP, MES, and QMS for manufacturing, and the README calls it the open core for manufacturing. The README feature checklist marks ERP, MES, QMS, custom fields, nested bills of materials, traceability, MRP, a configurator, MCP client/server support, an API, webhooks, and accounting as implemented. Capacity planning and simulation are on that same checklist as not yet done. These are maintainer claims from the repository documentation, not results of independent testing.

Who should consider Carbon

Carbon fits a specific profile, and it is worth a serious look only when that profile matches your operation:

• You do complex assembly, contract manufacturing, or configure-to-order work, where nested BoM, traceability, and a configurator matter.
• Your team is comfortable with TypeScript, React, Supabase or Postgres, and Docker. The README setup assumes all of these.
• You want an API-first core you can extend with custom apps. The README says the platform is designed to be extended through its API and points to an examples folder.
• You are interested in MCP client/server support for AI tooling, which is unusual to see listed in a manufacturing ERP.

If that profile fits, Carbon is most defensible as a foundation for custom development and a pilot, not as a drop-in replacement for a commercial ERP.

Maintenance signals

The repository is active. It was created in June 2024, and the last push was August 10, 2026, the same day as this review snapshot. At the snapshot it had 2,315 stars, 331 forks, and 144 open issues. That activity indicates interest and ongoing development, but it is not evidence of stability.

The biggest maintenance question is the absence of tagged releases in the collected data. Without a release history, you cannot judge the release cadence, evaluate a breaking-change policy, or pin a version. The README also documents recent internal migrations, including Trigger.dev to Inngest, Upstash to a standard Redis instance, and the Supabase CLI to a Docker Compose stack, which suggests the architecture is still evolving quickly.

Development sits under the crbnos organization, so scope, licensing, and roadmap decisions rest with that maintainer team rather than a broad community governance model.

Deployment concerns

Self-hosting is not a small install. According to the README, the local stack boots Postgres 15, GoTrue, Kong, Storage, Realtime, Studio, Inngest, Inbucket, and the edge runtime through Docker Compose, and the production notes point to Supabase components, Redis, and external services including Posthog for analytics, Stripe for billing, and Resend for email. Each of those is something you have to run, secure, and pay for.

Windows support is indirect. The dev CLI and setup scripts are POSIX-only, and the README says native cmd.exe and PowerShell are not supported; you need WSL or Git Bash. That matters for mixed-platform teams.

The optional Rust geometry service, which provides STEP to GLB conversion and assembly motion planning over FCL and OpenCASCADE, requires a native build with a patched OpenCASCADE that the README says can take 15 to 30 minutes per machine. ERP and MES run without it, so treat it as an optional add-on only if you need the 3D endpoints.

The evidence does not demonstrate production maturity. There is no documented production reference, benchmark, or security review in the snapshot, so verify those claims before committing.

Limitations and unknowns

• License ambiguity: GitHub reports Other (NOASSERTION), while the README says a private repository requires a commercial license to comply with the AGPL license. Get the exact terms reviewed before production, especially for private hosting or commercial redistribution.
• No tagged releases: upgrade paths and semantic versioning are unknown from this snapshot.
• Open issues: 144 were open at the snapshot, and without triage data this review cannot say whether they are bugs, feature requests, or stalled work.
• External dependencies: Stripe, Resend, and Posthog are part of the configured stack, so billing, email, and analytics are not fully self-contained.
• Unfinished features: capacity planning and simulation are unchecked in the README, which matters if you need production scheduling depth.
• No independent verification: this review is based only on the repository and its README.

Alternatives and evaluation questions

Do not choose Carbon simply because it is open source. Compare it against commercial manufacturing ERPs and other open-source ERP or MES suites using your own process requirements, and decide early whether a Supabase and Postgres stack fits the people who will operate it.

Before production, answer these questions:

1. What does the license permit for your hosting model, and who will review it?
2. What is the release and upgrade policy, and how do you pin a version?
3. Who owns the data, and how do you export it if you leave?
4. Which external services, such as Stripe, Resend, or Posthog, can you replace, and what does that cost in custom code?
5. Does the current feature set cover your core process, or would you be building MRP, configurator, or traceability behavior yourself?
6. Who will maintain the fork or custom extensions over three to five years?

Verdict

Carbon is a promising open-source foundation for manufacturers in its stated niche that already have TypeScript and Supabase capability. As of August 10, 2026, it is not a proven, release-managed product: the license needs review, releases are absent from the snapshot, and the deployment footprint is large. Evaluate it in a pilot with a small, well-scoped process, and budget for custom development and long-term maintenance instead of assuming it is turnkey.

Sources

This article is based on the public evidence listed below. Repository facts are attributed to the repository and its maintainers; no runtime testing was performed.

• crbnos/carbon repository on GitHub
• Carbon website (carbon.ms)
• Carbon documentation (docs.carbon.ms)
• Carbon examples folder
• Carbon public roadmap

If you need help deciding whether Carbon fits your operation, I can help scope a pilot, review the licensing and deployment trade-offs, or plan custom extensions around your actual processes. I am Xiang Peng at XP812, and I work on custom business software and legacy system modernization.