Bring Claude, Codex, Cursor or your own model. Define the system. DevMatrix deterministically compiles it into production code, tests, migrations, infrastructure and evidence.
Free · BYOK · Python/FastAPI · You own the code
Bring your own model key for authoring. Compilation itself calls no model — that's the point.
spec Ledger { module Accounts { entity Account { tenant_id: TenantId balance: Money @currency(ISO4217) } api REST { auth: JWT @multi_domain } } module Payments { flow Transfer { saga: orchestration steps: [debit, credit] } } compile { targets: [python_fastapi, java_spring] db: postgresql @rls, secrets: vault }}Nothing here asks for trust. What is published comes with the hashes to check it, and what is not published yet says so on the card.
Ask an assistant for a system twice and you get two systems, neither of which you can reproduce. DevMatrix does not compete with that layer — it consumes it. Your model decides what the system is; the compiler turns that decision into the same code, tests, migrations and infrastructure on every run.
Probabilistic · creative
Deterministic · auditable
AI reasons. DevMatrix compiles.
Your key authors the specification. Past the wall the compiler runs on its own and calls no model at all — which is the only reason the same specification can produce the same artifacts twice.
Four stages. You write the first one; past the boundary the compiler calls no model at all.
Your team declares the system in DMX — by hand, or with the LLM they already use.
Past the Compiler Wall there is no randomness. Same spec, same artifacts.
250+ quality gates across BLOCK, WARN and INFO, then 1,000+ generated tests per service. Nothing ships around them.
Production code, infrastructure, tests and a signed evidence pack — from one build.
Requirement changes? Edit the specification and recompile. The fix isn't a patch — it's a new baseline, identical for every service.
This is the source map the compiler signs into every build: every artifact traces back to the node that caused it.
Double-entry accounts, transfers and settlement. Four nodes, and everything they cause.
Illustrative DMX shapes, not customer builds. Published figures come from the verified build below.
Compliance asks for the personal-data column to be encrypted at rest. Here is the entire change.
spec Ledger { module Accounts { entity Account { tenant_id: TenantId pii: Json } } }The column is emitted encrypted, with a forward and rollback migration — in every service that touches the field.
Envelope keys are wired to the secret store the spec already declares. No credential reaches the source tree.
API contracts and serializers are regenerated so the field never leaves the perimeter in clear text.
New validation and security tests for the field are generated alongside the code, not filed in a backlog.
Threat model, SBOM and replay certificate are regenerated and signed against the new specification.
Nobody has to remember to do any of it, and nobody can do half of it.
A compiler that produced code you could not leave with would be a hosting bill with extra steps.
Free compiles complete Python systems — monoliths, with the same engineering quality as everything else we compile. Microservices, additional targets and advanced capabilities are paid.
Free limits the architecture, not the quality.
The specification is the input to the compiler, not a form on our site. Author it in the browser, in your editor, or from the agent you already have open.
/plugin marketplace add devmatrix-ai/devmatrix-mcp
/plugin install devmatrix@devmatrixInstall the DevMatrix plugin and your key is stored in the OS keychain — no plaintext config.
In most organizations that opens weeks of archaeology. Here the answer is a file: every emitted line traces to its specification node, and any third party can rebuild the artifact from the specification plus the compiler version.
Cloud SaaS · Self-hosted · Air-gapped · Hybrid. The hosted console is where DevMatrix runs unless your security boundary says otherwise.
How each one worksRequest beta access and tell us about the system you want to build. We’ll review your use case and get in touch.