Up and running in minutes

You'll have InTouch AI running on your own machine by the end of this page. No Docker, no cloud account, no infrastructure project.

Self-contained Java 17+ Self-hosted
Download InTouch AI Start at Step 1
Step 1

Before you start

InTouch AI is a single self-contained server that hosts both the API and the web UI — it runs anywhere Java runs. You need three things:

  • Java 17 or newer. Check with java -version. If you need it, grab the free Adoptium Temurin build.
  • Python 3 for the installer (one script, runs once).
  • An AI provider account. We recommend Anthropic Claude (the assistant is tuned for it), or Ollama to run fully local and free. Details in Step 4.

OS-wise: Linux, macOS, and Windows 10/11 are all fine.

Step 2

Download

One zip file, about 200 MB. It contains the server, the UI, the installer, and a handful of example skills and workflows to get you oriented.

Download InTouch AI
Step 3

Install

Unpack the zip and run the installer. The installer asks where to install and which port to use; the defaults are fine for a first try.

unzip install-intouch-ai.zip -d install-intouch-ai
cd install-intouch-ai
python install.py

The installer drops everything under the install directory you choose. Nothing is written outside it.

Step 4

Get an AI provider key

In InTouch AI, AI isn't a feature bolted onto an older core — it is the architecture. The vault, the scheduler, the connectors, the audit trail all sit behind it. So the AI provider isn't an add-on; it's the engine. The assistant drives InTouch AI by reasoning over a large context, so it needs a provider with a generous context window. We recommend Anthropic's Claude — it's what the assistant is tuned for. Want zero-cost and fully local? Ollama runs on your own machine, no key required — and the assistant runs very well on a local open model like qwen2.5: private, nothing leaves your network, zero per-call cost.

Recommended: Anthropic Claude

  1. Sign up at console.anthropic.com.
  2. Go to API Keys → Create Key. Name it "InTouch AI" and create.
  3. Copy the key — it starts with sk-ant-… and you only see it once. Keep it handy for the next steps.

Other strong choices

Step 5

Log into InTouch AI (for non-personal editions)

The installer already started InTouch, so open http://localhost:2200 in your browser. On the Personal edition you're a single user with fixed defaults, so you go straight in:

  • Username: intouch
  • Password: blueisle

These are intentionally immutable on Personal — a single-user, friction-free model. The Team, Department, and Enterprise editions add real user accounts, so on those you log in with the credentials your administrator issues — that's also where multi-user, access control, and password rotation live.

Step 6

Add your AI provider credential

This is the trust floor. Credentials live in an encrypted vault — workflows reference them by name, never pasted into scripts, never exposed even to the AI itself. The key you add here is locked down before it does a single thing.

  1. Sidebar → Credentials → Add
  2. Type: anthropic (or openai / gemini / ollama if you chose those in Step 4)
  3. Name: anything — My Claude Key is fine
  4. Group: Default Credentials
  5. Secret: paste the key from Step 4
  6. Save
Step 7

Configure the AI Assistant

The AI Assistant has its own provider settings (separate from the credential you just added — the credential is for workflows/tasks; this is what powers the chat).

  1. Sidebar → System → Settings → AI Assistant
  2. Provider: Anthropic (or whatever you used in Step 4 and Step 6)
  3. API Key: paste the same key here (or your provider's key if you went with a different one)
  4. Model: pick from the list — the default for each provider is the recommended starting point
  5. Enabled: check the box
  6. Save
Step 8

Say hi

Open the AI Assistant chat (sidebar → the chat icon, or wherever the current build labels it). Type something simple:

What can you help me with?

You'll get a response in a moment. Try asking it to list your workflows, summarize a recent run, or build a new automation. This is how people work now: describe what you want, don't configure it. The assistant has tool-use access to your InTouch AI server — it can read state and take actions on your behalf. And when a workflow fails, it doesn't hand you a stack trace and a retry counter — InTouch AI reads the failure, tells you why, and fixes what InTouch AI can. "It broke. Here's why. I fixed it." No older tool can say that.

Step 9

Browse the InTouch Hub

Find it and run it. The InTouch Hub is a library of ready-to-run skills and workflows you install with one click — start from something that works instead of a blank page. Browse from the sidebar (Skills view or Hub browser) or visit hub.blueisle.com directly. And because every InTouch artifact is a single self-contained file — an IML tool, workflow, or monitor, or a native SKILL.md skill — anything you build, you can hand straight to another InTouch user who uploads it to their instance, peer-to-peer, no Hub required.

The items are example starting points — provided as-is for you to install, read, and adapt to your setup. They're not productized templates that will work out of the box for everyone. Read each one's README before running.

For ideas grouped by use case, see the Examples by vertical on the home page.

Starting InTouch AI after your first session

The installer starts InTouch for you the first time. When you come back later — after a reboot, or once you've stopped it — start it again from the Desktop shortcut, or from a terminal:

cd ~/intouch-ai
python bin/start.py

It comes up in a couple of seconds. Then open http://localhost:2200 — InTouch AI is one self-contained server hosting both the REST API and the web UI, same origin, no separate UI to start.

Step 10

Got stuck?

We're happy to help — and your feedback makes the product better. If anything breaks, confuses, or could be improved:

Tell us what you were trying to do and what happened — screenshots help. Someone in the community (often us) will jump in.

Automate your first thing today

This is AI-native, not AI-retrofitted — production-hardened in Fortune 500 production at enterprise scale, before "AI automation" was even a phrase. Self-hosted, from your laptop to the enterprise. The hub content is illustrative starting points. Everything else is between you and your machine.

Download InTouch AI See example automations