01Independent of any one AI company
No. CodeHero works with many AI companies at once and with models on your own hardware. You can change company or model at any time — the project, its rules and its history stay.
Most AI coding tools belong to one AI company. Their models, their prices, their rules, their outages — and when they change, you change with them.
CodeHero is built the other way round. The factory is yours; the AI models are workers it hires. It works with Anthropic (Claude), OpenAI (GPT and Codex), Google Gemini, xAI Grok, GLM, Qwen, OpenRouter’s catalogue, and local models through Ollama or vLLM — by API key or by subscription.
It also uses them together: one company’s model builds, another company’s model checks. A table of the available models helps choose the right one for each job by ability and cost.
If a company raises its prices, slows down or disappears, you switch. Nothing is lost, because the specification, the decisions, the code and the history live in your installation, not inside an AI company’s product.
↑ Back to the list02Why your secrets must never reach the AI company
No. Passwords and keys you store as secrets are used by your application, but their values are never sent to the AI service.
An AI model works on text. Whatever it is shown — a file, a command, the output of a command — leaves your machine and goes to the AI company’s servers. If a database password or a payment key is in that text, it has left too: into logs, into history, possibly into places you will never see.
With ordinary tools this happens quietly, every day: a configuration file is read, a connection string is printed, a key ends up in a prompt.
CodeHero stops it inside the factory. Passwords and keys you store as secrets are used by your application, but their values never reach the AI: a command that would carry one is refused, and any output that contains one is replaced before the AI sees it. They are also kept out of the project’s git and ZIP backups by default.
One honest limit: it protects what you store as a secret (eight characters or longer). A password typed into a chat message or visible in a screenshot cannot be recognised — so store every password as a secret.
↑ Back to the list03Many AIs together make fewer mistakes
Because one AI does not catch its own blind spots. In CodeHero one model builds and a model from a different company checks; problems go back for repair and are checked again.
Every AI model has blind spots — and it tends not to see its own. Asking the same model “is this correct?” usually gets a confident yes.
CodeHero works like a good team instead. One model builds. A model from a different AI company checks the work against the agreed rules and records what passed and what failed. It does not change the code; it opens a repair. The repair is checked again, and repeated failure stops for a person’s decision instead of looping.
For harder plans a second model reviews the plan before the work starts, and larger work closes each phase with a review before the next phase begins.
You can see this in our recorded example: the builder finished, the independent check found two problems on phones that the builder had missed, they were repaired, and the recheck passed. One AI alone would have delivered them.
↑ Back to the list04Not another coding tool — a category of its own
It is a different kind of thing. Coding tools help a person who writes code. CodeHero does the whole job — asks, plans, builds, checks, hosts and maintains — for the person who knows the business.
AI coding tools are made for programmers. They sit in an editor and help someone who already knows how to write, test, deploy and maintain software. Comparing CodeHero with them is like comparing a factory with a power drill.
CodeHero is made for the person who knows the work: the owner of a repair shop, a clinic manager, a business that needs a platform. That person does not write code and does not supervise each step. They answer the questions only they can answer, approve a written plan, and receive working software.
Between those two moments, the factory does the whole job: it plans in phases, assigns each task to a suitable AI model, builds, has a different model check, repairs, hosts the application with its domain and backups, and returns to the same project for every later change.
Coding tools can be one of the workers inside it — Claude Code, Codex and Grok Build run as workers in the factory. That is the difference: they are tools; CodeHero is the place where the tools work.
↑ Back to the list05CodeHero is the developer. You are the advisor.
You become the advisor: you decide the business rules, approve the plan and accept the result. CodeHero is the developer that asks, builds, checks and maintains.
For decades, getting software meant explaining your business to developers — and then explaining it again, and checking their work, and waiting.
CodeHero turns this round. It plays the developer’s part: it asks, writes the specification, plans, builds, tests, repairs and maintains. You play the part that only you can play: the advisor who knows how the business really works.
You are asked about what matters — who approves, what happens at a limit, what must be kept and for how long. You see the written plan and say yes before anything is built. A missing business decision comes back to you rather than being guessed. At the end you receive the working result, its checks, and a clear list of anything still open.
You do not need to read code, watch every step or learn a tool. Your knowledge of the business is the input the factory cannot do without.
↑ Back to the list06How AI memory works — and what memory folding changed
With memory folding: when the AI has finished with a long result it keeps a short note of what matters and puts the rest away — it can open it again at any time. So it works much longer without losing the thread.
An AI model has a working memory of limited size: everything it has read and done in a job must fit in it. On long jobs it fills up. The usual remedy is to summarise everything at once — and the summary loses details the work still needed. That is when an AI starts repeating mistakes or forgetting decisions.
CodeHero added memory folding. Think of a person at a desk: when they finish with a long report, they write a short note of what matters and put the report away in a drawer. The desk stays clear, and the report is still there.
The agent does the same with its own work: a long file listing, a test log, a big output it no longer needs in full becomes a short tag with its note — the file and lines, the values, the next step. The original is kept and can be opened again at any moment. Pictures it has already looked at are put away automatically (when CodeHero talks to the AI directly), and its reasoning is kept between steps.
The result: agents work much longer on the same job without losing the thread, and nothing they saw is ever really lost. It works with every AI the factory uses.
↑ Back to the list07Real economy: an AI that does not forget or get confused
The factory gives it structure: smart tickets with one clear job each, a change ledger of every decision, a map of the code and search by meaning. The AI reads what it needs instead of guessing — which also saves a lot of cost.
On a big project an AI wastes most of its effort finding things: which file does what, what was decided last week, what changed since. When it guesses, it gets confused, contradicts earlier work and has to be corrected — and every correction costs time and money.
CodeHero gives the AI the structure a well-run engineering team has. Smart tickets: each job is a ticket with one clear objective, the rules it must keep and how to check it is done — and the platform loads the right guide for the kind of job. A change ledger: every change you ask for after the plan is recorded with its reason, and every later ticket reads it. A codebase map: the AI asks where a function or table is instead of reading the whole project. Search by meaning: it finds what was said, decided or built — even when different words were used.
There is also a knowledge base for each project, where decisions and lessons are stored and found again.
The AI reads exactly what it needs, at the moment it needs it. That is why it stays consistent on long projects — and why it uses far fewer tokens than an AI that rereads everything.
↑ Back to the list08Why a software factory, not a tool
Because it runs the whole production line on its own — planning, building, independent checking, repair, delivery and maintenance — with stations, rules and records, like a factory. A tool waits for a person to use it.
A tool waits for a skilled person to pick it up. Its result depends on that person’s skill and attention, every time.
A factory is different: it has stations, procedures and quality control, and it produces a reliable result again and again. That is what CodeHero is. The order comes in as your answers. Planning turns it into phases and tickets. Each station is staffed by a suitable AI model. Quality control is a different model that checks the work against the agreed rules. Faults go back to repair. The finished product is delivered, hosted and maintained.
And like a real factory, it keeps records: every decision, every change, every check and its result, on your machine.
This is why the person who knows the business can get software without becoming a programmer — and why the same quality comes out of the factory on the tenth project as on the first.
↑ Back to the list09How it can change the world of applications
When the person who knows the work can get software made, checked, hosted and maintained without a development team, software can be made for every business and every need — not only for those that can afford a team.
Today, custom software is a privilege. A small business, a clinic or a workshop usually makes do with a generic product, or with spreadsheets, because a development team is too expensive and too slow for them.
Most of the knowledge needed for good software is not programming. It is how the business works: who approves, what the exceptions are, what must never be lost. That knowledge belongs to the people in the business — and until now it had to pass through developers to become software.
CodeHero lets it go straight into software. The person who knows the work answers the questions, approves the plan, and the factory builds, checks, hosts and maintains. Changes come back to the same factory whenever the business changes.
When that becomes ordinary, software stops being something you buy and adapt to, and becomes something made for you — for one shop, one clinic, one way of working. That is the change we are building for.
↑ Back to the list10A better working life for developers
Less repetitive work and fewer late nights. The factory takes the routine building, checking and fixing; developers decide the architecture, review what matters and handle the hard problems.
Much of a developer’s week is not the interesting part: the same forms and tables again, chasing a bug that appears only on phones, rerunning tests, explaining the same rule twice, fixing something late at night before a release.
In CodeHero a developer works like a lead engineer with a tireless team. The factory takes the routine: building to an agreed specification, independent checking, repair rounds, installation and upgrade checks. Every result arrives with its checks and a clear list of what is still open, so review is quick and focused.
The developer keeps the parts that need a person: the architecture, the hard decisions, the review of what matters, the conversation with the client. They can open any ticket, read its reasoning, send it a message or stop it.
Fewer emergencies, less repetition, more time for the work that needs experience. That is not a smaller job — it is a better one.
↑ Back to the list11Why it is a hosting platform too
Because building and running belong together. The application runs where it was built, with its domain, secure connection, backups and recovery — and every later change returns to the same factory. No separate hosting panel to learn.
Usually, building software and running it are two different worlds: one team builds, then the result is moved to a hosting service with its own panel, its own settings and its own surprises. Every later change has to cross that gap again.
CodeHero keeps both in one place. The application runs where it was built — directly on the machine or in its own isolated container — with its domain, secure connection (SSL), databases, backups and restore. It works like a control panel for every language and database the factory uses, so there is no separate hosting tool to install and learn.
Because the factory knows exactly how the application was built, it also knows how to run, back up, upgrade and repair it. A change request goes into the same project, is built and checked, and goes live from the same place.
One place, one record, one responsibility — from the first question to the hundredth change.
↑ Back to the list12Always open-source tools — the future of technology
Always open, widely used technologies — PHP and MySQL first, and others such as Node, Python or .NET when the job needs them. Your application is never locked into a proprietary platform.
Software built on a closed, proprietary platform belongs, in practice, to that platform. If its prices or terms change, the application has to follow — or be rewritten.
CodeHero builds on open, widely used technology. The platform itself runs on Ubuntu with MySQL, nginx, PHP and Python. The applications it builds use open stacks — PHP and MySQL first, because they are simple, fast and run everywhere, and Node, Python, .NET or others when a project really needs them.
That means any developer can read, change and maintain what the factory built, on any ordinary server, without CodeHero. The code is plain code, in your project folder and your git history.
Open source is where the world’s software infrastructure already lives, and where it is going. Building on it keeps your application yours, for as long as you need it.
↑ Back to the list13Why subscription and not pay-per-use
Because in our own measurements, for heavy daily work, a subscription has cost up to about 40 times less than paying per use for the same work. CodeHero runs the subscription AI tools (Claude Code, Codex, Grok Build) as workers in the factory.
A software factory uses a lot of AI: planning, building, checking, repairing, rechecking — all day. Paid per use (per token), that adds up quickly, and the bill is hard to predict.
AI companies also sell subscriptions at a fixed monthly price, made for their own coding tools. In our own measurements, for heavy daily work like a factory’s, a subscription has cost up to about 40 times less than paying per use for the same work. How we measured: we counted the tokens the factory used until a weekly subscription limit was reached, and priced the same tokens at the pay-per-use rates of the same model.
CodeHero is built to use them. It runs Claude Code, Codex and Grok Build as workers inside the factory, signed in to your subscription, and it also supports coding plans such as GLM’s and QwenCloud’s. When choosing which model does a job, the factory counts a subscription as much cheaper than per-use billing, and it never silently falls back from a subscription to paid-per-use billing.
You can still use API keys wherever they make sense. The point is that the economics of the factory are under your control — a fixed cost you know, instead of a bill that grows with every step.
↑ Back to the list14Why the knowledge is built — and stays — inside the box
It is built up and kept inside your installation: decisions, rules, changes and their reasons, next to the code. A new person or a new AI continues from there — the knowledge does not leave with them.
In most projects the most valuable thing is not the code. It is the knowledge around it: why a rule exists, which exception the client asked for, what was tried and failed. Usually that knowledge lives in people’s heads, in chat threads and in emails — and it leaves when they leave.
In CodeHero that knowledge is written down as the work happens, and kept with the project on your machine: the agreed specification, every decision and its reason, every change you asked for after the plan, every check and its result, and a knowledge base the agents add to and search.
So nothing depends on one person or one AI. A new developer, a new AI company or a new model can read what was decided and why, and continue from there. The next change starts from the whole history, not from a guess.
And because it lives in your installation, it is yours — not stored in an AI company’s product, not lost when a subscription ends.
↑ Back to the list15How small local AI models can work like big ones inside the factory
Inside the factory, yes — more than on their own. Small, clear tickets, a checklist for finishing, tool help and a stronger model that checks the result let small models on your own hardware do much of the building.
Small AI models that run on your own hardware are private and cheap to run — but on their own they lose track on long tasks, skip steps and declare a job done too early.
Inside CodeHero they do not work on their own. The factory gives them what a junior engineer gets in a good team: a small, clear ticket with one objective; a task card and a completion check that does not let the job end before what was asked is done; tool help that tells them when a file changed since they read it; memory folding to keep their working memory clear; and the right guide for the kind of job.
Then a stronger model — often from another company — checks the result against the agreed rules and sends problems back for repair. The small model does the volume of work; the big model supervises where judgement matters.
Our goal is for local models to write most of the code, with big models planning and checking. We are measuring models on real factory tasks, so that the factory can give each job to the cheapest model that does it correctly.
↑ Back to the list 
CodeHero