About Codee3
Codee3 installs the structure, the routing and the guards. It does not author your knowledge, and it never will.
That is the most expensive decision on this page and the one that explains the rest of it. A governed layer full of confident, unverified generated prose would be worse than no layer at all — so the product declines to produce it, and says so on every page where somebody might hope otherwise.
Why Codee3 exists
A software repository is exceptionally good at storing implementation. It has never been good at storing the organisational knowledge about that implementation.
For decades, humans bridged the gap. We kept the context in our heads, in chat histories and in wikis. When we read code, we silently applied constraints that were never written down anywhere.
AI coding agents exposed that gap and amplified it. An agent reading a repository has no intuition. It takes a stale plan at face value. It cannot tell an adopted decision from an abandoned proposal. It has none of the institutional memory a long-serving engineer carries around unexamined.
Codee3 was built to give that memory a governed place to live — beside the code, in a format both humans and agents can navigate reliably, guarded against structural decay.
Product principles
Six commitments that decide what gets built and, more often, what does not.
- Local-first knowledge
- The repository is the source of truth for implementation; it should also be the source of truth for the institutional memory surrounding that implementation. Knowledge that lives elsewhere cannot be reasoned over reliably.
- Plain files over platform dependency
- Governance should not require a SaaS platform. A governed structure of plain Markdown and zero-dependency Node scripts outlasts any startup, and keeps your knowledge readable without proprietary tools.
- Authority must be declared
- Nothing can distinguish a locked decision from a discarded proposal unless the document itself declares its status. Authority has to be explicit, not inferred from context that no longer exists.
- History belongs in lineage, not filenames
- Appending -v2 or -FINAL to a filename forks truth. A document needs one stable identity; its evolution belongs in a changelog, not in its title.
- Derived summaries never become authority
- Documentation is a projection of reality. Where a summary contradicts the implementation or the canonical decisions, the summary is the thing that is wrong. Projections must be attributed and their staleness must be visible.
- Technical proof is stronger than marketing proof
- We do not claim Codee3 solves every agent problem. There is no fabricated social proof and no invented metric. We publish the actual terminal output of the tools and let it be checked.
What Codee3 is today
Present tense only. Nothing in this section describes a roadmap item as though it already shipped.
- Current technical delivery
- Non-destructive scaffolding. 57 plain files. 14 zero-dependency npm scripts. Agent surface adoption across ten tools. Five guards wired into CI. Read-only, deterministic repository intelligence.
- Explicit authoring boundary
- Codee3 installs the structure, the routing and the structural guards. It does not author your knowledge for you — you and your agents author the content.
- Current commercial status
- Pre-launch. Commercial access is not open yet. The intended model is a one-time licence per legal entity.
- What is planned
- No hosted service. No telemetry. No live-syncing tier. Codee3 will remain repository-native.
Who is behind it
Jeremy Linton, Founder & CEO
I got tired of explaining the same decision to an AI for the fifth time.
At first, AI-assisted development felt fast. Then the projects grew. Agents reopened settled questions, re-read the same APIs, lost requirements that had already been stated, and built fixes for the code in front of them instead of the system we were actually trying to create.
I tried better prompts. Longer plans. More context. None of it fixed the underlying problem.
The agents could read the repository. The repository could not tell them what was true.
A locked architectural decision, an old plan and a discarded proposal could all look equally valid. Once I saw that, the problem stopped looking like agent memory and started looking like knowledge governance.
Codee3 is the system I built in response. It installs an authority-graded knowledge layer beside the codebase, separates governed decisions from drafts and superseded material, directs agents toward the right context, and enforces the structure through deterministic checks in CI.
I have now used it across more than five projects. The goal is simple: stop making the developer repeatedly remember on behalf of the machine.
Millie Newall, Co-Founder & Chief Strategy Officer
Leading company strategy, market positioning, partnerships, AI efficiency and environmental responsibility.
Millie Newall is Co-Founder and Chief Strategy Officer at Codee3, where she leads the company’s strategic direction and helps ensure its approach to AI-assisted development is efficient, commercially relevant and aligned with the company’s broader vision.
With a background in healthcare, clinical education and research, Millie has experienced firsthand how artificial intelligence can transform the way people learn, work and solve complex problems. She has applied AI to analyse health research, improve access to evidence, support women’s health initiatives and develop tools that make complex information more practical and accessible.
As AI became more capable, she also recognised a growing challenge. Too often, developers and organisations spend valuable time re-prompting models, correcting inconsistent outputs and rebuilding lost context. Beyond affecting productivity, this unnecessary repetition increases computational demand, costs and energy consumption.
Millie co-founded Codee3 to help address this problem.
Codee3 enables AI to work with persistent project knowledge, structured rules and shared context, helping development teams build software more consistently and efficiently. By reducing avoidable rework and improving the reliability of AI-assisted development, Codee3 helps organisations gain greater value from AI while minimising unnecessary compute and waste.
Millie is passionate about improving how AI is used in practice. She believes the next stage of AI is not only greater capability, but greater efficiency — for businesses and the planet.
Where to go from here
- How it worksThe installation, the structure it writes, and the checks that hold it together.
- The architectural modelHow authority flows, how a fact earns standing, and what the model deliberately excludes.
- Why repositories are hostile to AI agentsThe failure modes this product was built against, argued at length.
Commercial access is not open yet. The access page sets out the intended terms, in full, before there is anything to sell.
Intended access terms →