About us
Software should do what you ask of it
Syncanix builds the Experience Agent, one script that turns an interactive product into something its users can operate by asking. This page is the idea behind it, the rules we hold it to, and a straight account of how far along it is.
The idea
After UX comes AX
Every serious product now holds more capability than any one person will ever find. The industry’s answer has been UX: menus, tours, tooltips, documentation, support queues. Decades of real craft, and every bit of it aimed at helping people find things. None of it does the thing for them.
AI changed what an interface can be. When software can understand a request and carry it out safely, the honest interface is no longer a better menu. It is a conversation backed by real capability. Not a chatbot that quotes the manual, but an agent that does the work, signed in as the person asking, holding exactly their permissions and nothing more.
We call that discipline Agentic Experience, or AX. Where UX arranges a product so a person can operate it, AX puts an agent inside the product to operate it for them, and to ask before anything changes. Syncanix exists to make that the default for the interactive web. Not a rewrite, not an integration project: one script tag, and a product that reads itself.
- 1
- script tag to install
- 39
- frameworks discovery reads
- 6
- languages, right-to-left included
- 100%
- of actions in the audit trail
What we build
One agent, everywhere your product is used
The Experience Agent installs once, learns your product on its own, and then answers and acts for the people using it. On your pages, and in the AI tools they already have open.
It reads the product itself
Discovery looks at your running product and your code, across 39 frameworks in 19 languages, and builds the catalog of what the agent may do. There is no spec to write, and the catalog is checked against your product so it cannot quietly go stale.
It answers, then it acts
Ask what something means and it explains, grounded in your own docs. Ask for the thing itself and it does it: fills the form, makes the call, walks the flow. Always as the signed-in user, never as more.
It goes where your users already work
The same capabilities, the same permissions and the same audit trail are served to Claude, Cursor and ChatGPT through a per-workspace MCP server. Your product becomes operable from the tools people keep open all day.
It builds what the product never had
Charts, summaries, comparisons, guided walkthroughs, composed on the spot from live answers, so an unusual question gets a working interface rather than a paragraph.
The rules
Five things it will never do
Asking to act inside your product is a serious request, so the limits are designed in rather than promised. Every one of them is observable in your own audit trail.
It never acts without asking
Anything that changes data shows the person the exact change and waits for them. On your own pages there is no exception at all. In an AI client, a workspace can pre-approve routine writes it has reviewed itself, and even then, never a deletion, and never a capability the agent found on its own.
It never outranks the person using it
The agent acts as the signed-in user and inherits their permissions exactly. There is no service account behind it with a wider view of your data.
It never moves your data, or trains on it
Customer data is stored in Frankfurt and the AI runs in the same region. Nothing you or your users send is used to train a model, ours or a provider’s.
It never leaves an action unrecorded
Every action lands in an audit trail you can read and export. The trail is tamper-evident: a row altered after the fact shows up as altered.
It never treats a language as an afterthought
Six languages, with Arabic and Hebrew laid out right-to-left, are part of what finished means here: the interface, the agent’s replies, and the instructions behind them.
Straight answers
Where it actually stands
Nothing on this site is described as finished unless it is running in production today. Here is the line.
- Running in production
- The agent on your own pages, the Claude, Cursor and ChatGPT connection, discovery from your live product and code, the approval and audit layer, and all six languages.
- Still ahead
- SOC 2 is on the roadmap rather than in hand, and some compliance documents are drafted but awaiting counsel review. Where that is the case, the Trust Center says so on the document itself rather than in a footnote.
- Talking to us
- Prices are published, so you can size it before speaking to anybody. When you do reach out, a person answers, and the security question, or the thing you are sure will not work in your product, is the conversation we would rather have first.
Every claim on this page has a document behind it in the Trust Center, including the ones that are still open.
Make your product an agentic experience
The fastest way to judge this is to point the agent at your own product and ask for something real. The second fastest is to ask us the question you think it cannot answer.