Your AI works like an engineer.
SonIQ scans your codebase into living documentation — the structure is extracted from code, not generated by an LLM. Your AI adds the meaning and works through a real backlog: tasks, reviews, history. You see everything on the board.
Needs Node.js 20+. macOS and Linux; on Windows, use WSL.
Memory is private. Documentation is shared.
Every agent keeps its own opaque memory — unreadable, unverified, gone tomorrow. Teams don't run on memory; they run on documentation, backlogs, and review.
Without SonIQ
Each agent hoards its own blob. You can't read it, teammates can't share it, nobody verifies it.
With SonIQ
One living, code-grounded doc set — readable by you, shared by every agent, verified against the repo.
Scan once. Connect via MCP. Work.
Scan
The scanner extracts structure from code — endpoints, schemas, components. Deterministic: no LLM, zero tokens.
Meaning
Your AI augments the docs with the why — every write passes the gates; cited file-and-line spans are checked, uncited claims are marked.
Work
Agents file tasks, claim them, ship through review — a real backlog you see on the board.
Built for code, not chat.
Extracted, not generated
The structural layer is pulled from code by a deterministic scanner, not written by an LLM — and the coverage report says what it doesn't cover.
Docs checked against the code
Drift is flagged. Docs for deleted code are retired. Changes are attributed.
A real backlog for AI work
Tasks, triage, review — agents work like engineers, and you stay in control.
Kept in step, vendor-neutral
The soniq daemon sends your project's files to SonIQ, where they are scanned — like a git host, for knowledge. Saves refresh the code index within seconds and most knowledge entries after a short quiet period; a full soniq scan refreshes the rest. Works with any MCP client.
Give your AI a real place to work.
Free forever for the structural layer. Pro, for when your agents start writing, is coming soon. No card to start.