InTouch AI vs n8n — Comparison Sheet

AI is the architecture, not a feature. InTouch AI was built with AI at the center — the credential vault, the 11 database drivers, scheduling, RBAC, and audit all sit behind it. n8n bolts an AI Agent node onto a node-graph core that predates it. A general AI-native engine can do everything a node-and-connector tool does; the reverse is impossible — n8n cannot grow an AI core after the fact.

InTouch AI, no spin: self-hosted, free Personal edition — genuinely free, no workflow-count cap, no feature gates; single-tenant by design; runs behind NAT via Tailscale; a real governance layer (encrypted credentials, per-object RBAC, full audit); agentic AI built in. Against n8n: n8n's self-host runs free only up to 20 active workflows. InTouch AI's Personal edition is free forever. And InTouch AI talks to 11 databases for real — Oracle, DB2, Essbase, TM1, enterprise systems — not HTTP nodes pretending to be a database driver.

The contract, now intelligent. Tell InTouch AI what to do, when, what to do when it works, what to do when it doesn't, and who to notify. In the config era the "doesn't work" clause was a dumb rule — retry N times, email a log. InTouch AI turned it into an assessment: it reads the failure, knows why, smart-retries, refreshes an expired token, and surfaces the one sentence that matters. "It broke. Here's why. I fixed it." A node-graph can re-run a failed branch; it cannot tell you why the branch failed. And the determinism dial: run a workflow as pure AI, then graduate it to fully deterministic, zero-AI-cost, identical-every-time, fully-audited as it earns your trust. n8n gives you one fixed point on that line. InTouch AI gives you the whole dial.

The credential vault is the trust floor. AES-256, referenced by name, never written into a workflow, never exposed in a script — and never exposed even to the AI itself. The intelligence runs on top of the vault; it never sees inside it. That is what lets you put an AI-native engine in front of Oracle, CyberArk, and production credentials without flinching.

Don't start from a blank canvas — find it and run it. The Hub is a library of ready-to-run tools, skills, and jobs: install, point it at your setup, run, in your own language. Start from something that works. And self-automation makes the move off n8n cheap — InTouch AI ships an intouch_publisher credential type that, when attached to a runtime env tool, gives a Python script a session token to call the InTouch REST API as a real publisher — with that publisher's RBAC and full audit. Write a script that reads your n8n workflow exports, builds equivalent InTouch AI jobs via REST, and runs once. No Java/Kotlin code, no server release, no native importer required. See self-automation.html.