Skip to content

CortexBuild

The model is a commodity; the harness is the moat.

CortexBuild is a model-agnostic, governed builder platform for org-scale software engineering on your existing repos. It is not a single prompt or a single model — it is the control loop, governance, and guardrails that render your team's rules into the AI coding tools you already use, so the agents you run can build on the systems you already have: new features, services, endpoints, and refactors, plus legacy modernisation and backlog burn-down. Brownfield is a featured strength, not a ceiling.

The open-source engine, cortexbuild-core, exists today and ships as a CLI plus committed artefacts under .cortexbuild/.

In plain English

CortexBuild reads your existing codebase, captures the unwritten conventions your team already follows as rules, and feeds those rules to every AI coding tool you use — so the agents stop breaking your standards. New here? Start with What is CortexBuild?

What the harness does

  • Rendered review pipeline. CortexBuild renders a fixed, auditable sequence of specialised review roles — Planner → Critic → Builder → Tester → Security → Optimizer → PR Reviewer — into the AI coding tools you already use, so every change those tools make is planned, reviewed, built, and verified before you choose to merge. CortexBuild renders and governs the workflow; your host AI tools execute it. See the Architecture Overview and System Architecture.
  • Rules as a single source of truth. Governance rules are authored once and rendered to every agent surface, so policy is consistent no matter which tool or model executes. See the Rule Packs outline.
  • Deterministic eval & guardrail gates. Conformance scoring, the offline eval harness, render --check, and recurrence-ledger guards are deterministic checks, not vibes — they flag a change the same way no matter which model produced it. CortexBuild flags failures; it does not enforce merges by itself — your existing approvals decide what lands before you choose to merge. The Critic, Tester, Security, and Optimizer roles add LLM review on top. See Conformance.
  • Recurrence ledger + learning loop. Defects and their fixes are recorded in a recurrence ledger so the same class of failure is caught earlier next time. See ROI & Insights.
  • Provider router + BYO-keys. A model-agnostic provider router lets you bring your own keys and swap providers without rewriting your harness. See the Provider Adapter and Pluggable Adapters.

Start here

Follow the journey top-down — understand → install → onboard → learn → build → scale → business case:

Looking for the product site? Visit cortexbuild.dev.