Not a feature list. The actual system map, annotated — read the tree and you have read the system. Plain folders and Markdown files on your disk. No account, no cloud, nothing to buy.
Every node below maps to a real folder or file in the Synta system. Expand a branch, open a node, and read what it is, why it exists, and how it works. Follow the links between related nodes — the same shape the system uses to think.
Synta is not another AI subscription. You run it on your machine with the AI tools you already use: Claude, Codex, Cursor, an Ollama local model, or a similar agent. Your subscription, API key, or local model stays yours. Your files stay yours.
Download or clone the public repo into a normal folder. Nothing starts in the cloud. The system begins as files you can open, move, back up, and inspect.
The dashboard gives non-technical users a web interface for the system: projects, tasks, health checks, knowledge graph, and agent sessions. More tech-savvy users can use the command line instead. There is nothing to be scared of: either way, the same files and rules drive the work.
An AI-led setup creates your DNA: profile, preferences, first projects, and the rules for how the system should work with you. After that, you learn by using it and feeding it real data.
Most people do not fail with AI because the model is weak. They fail because they skip the basic loop: context, goal, action, feedback, and conscious decisions. Synta makes that loop visible and repeatable.
Tell the AI who you are, what matters, how you work, and what world it is stepping into.
If the goal is unclear, use AI to make it precise before asking it to produce anything.
Do not overthink what AI might do. Give it context, run the session, and inspect the result.
AI is useful only when you act on the recommendation, test it, or turn it into a decision.
Correct it, refine it, and tell it what was useful. No feedback means no learning loop.
The worst failure mode is skimming the answer and letting the AI decide by accident. You own the decision.
Do not automate the fantasy workflow. First prove the pattern in live AI sessions, then automate it.
The map explains the structure. The tools make it useful every day: they help you plan, prepare, review, start focused sessions, and notice what you forgot to write down. A founder, manager, consultant, or developer will use different routines, but the underlying system stays the same: projects, context, tasks, agents, and decisions in one place.
Open projects, review tasks, inspect the knowledge graph, and start agent sessions from one web interface. With Tailscale, the dashboard can be reached from any trusted laptop, desktop, or phone.
Pulse sends a daily focus brief to Telegram: what changed, what needs attention, which tasks are stale, and where yesterday's work created new decisions or follow-ups.
Calendar, email, and Slack adapters can add a layer of automation: upcoming meetings, raw context, and routing hints arrive before you have to hunt for them manually.
Add tasks, plan them, review progress, and start an AI session from a card. The board is not just tracking work; it is a launchpad for making progress.
Synta is named for the way it works. Knowledge is organized in levels — areas hold projects, projects hold files. And it's connected by links — the system's synapses — so it can pull the right context and spot what's missing. The map above navigates exactly that shape.
Root, system, agents, areas, projects, files. Every level is a real folder you can open. Context loads by level: a handful of files on startup, the rest only when a task needs them.
Files link to related files. The system walks those links to gather context, and flags dead links, missing files, and contradictions on every read. In the map, every "see also" is one of these links.
It's all Markdown in folders you own. Open it in any editor, back it up like any normal folder, and use Git if you want version history. Walk away tomorrow and you keep every byte — because it was always just text.
The system is readable before you download it. Here is what you can verify for yourself.
The startup routine, the routing logic, the rule that stops the AI from inventing facts, and the limits on what it may change — none of it is hidden inside an app. Open the file and read the rule.
It runs as a folder on your machine. Data only leaves if you connect Claude, OpenAI, or another AI provider yourself. Synta does not send your files back to a server you do not control.
A public project with a standard open license. Use the whole system, or copy one useful rule into your own setup. It's built to be read and reused, not to trap you.
The format has been stable for twenty years. Whatever happens to Synta, your files still open in every text editor ever made.
I am Emil Majkowski, a CIO / systems architect and former founder.
I spent 15 years building both the business and the technology of a vacation-rental company from zero to acquisition. We were building infrastructure for a market that was still learning what it needed.
Today I work as a CIO / systems architect and spend most of my thinking time on AI: how models work, how people use them, where AI workflows break, and what it takes to make AI useful inside real companies.
Security is one of the main criteria in everything I build. That is one reason Synta is local-first: your memory, rules, and work context should stay under your control.
Synta is the personal operating system that came out of that work: local memory, clear rules, specialist roles, and a protocol that makes AI useful beyond a single chat.
I am sharing it because I believe AI creates real value only when it has context, boundaries, feedback, and a human making conscious decisions.
No price, no checkout, no fit-call. The map above shows the structure and explains the key parts; when the public repo is ready, you will be able to inspect the full prompts and files yourself.
The map is readable here today. The full system — prompts, rules, roles, protocols, and files — goes public shortly in the GitHub repo.
Read the map now; download it, run it, or just steal the ideas the moment the repo is up.