InTouch AI

Automation you describe in plain English. Tell the AI assistant what you want done; it builds the automation, schedules it, runs it, and shows you proof it ran. Your credentials stay in an encrypted vault the AI cannot read.

Personal Edition is free forever — no licence key, no trial, no expiry.

Install

Needs Java 17+. On Windows the installer installs Python for you.

Get install-intouch-ai.zip — from Releases or blueisle.com. It is the complete installation, ~240 MB: the server, the UI, the docs, the starter tools and workflows, and install.py, which copies them where you choose. Nothing is fetched over the network, so it installs on a machine with no internet.

GitHub's green Code ▸ Download ZIP button is not this. That button archives the repository, which holds only this README and the licence.

Windows — extract install-intouch-ai.zip, double-click install.bat

Linux / macOS

unzip install-intouch-ai.zip -d install-intouch-ai
cd install-intouch-ai
python3 install.py

Log in at http://localhost:2200 (intouch / blueisle), paste an openrouter.ai key into the Get Started panel, and type what you want.

Put a dollar on the key — a free key is good for a handful of calls and then stops. Or run a local Ollama model instead, with no key and no cost.

Start and stop with python3 bin/start.py and python3 bin/stop.py.

What's in it

Already using Claude Code?

Then you have the authoring half of the problem solved, and this is the other half.

An AI coding agent decides what should happen. It cannot make it keep happening — it cannot wake up at 03:00, wait four hours for a condition without burning context, do the identical thing on the hundredth run, or hold a production credential it has not read. Those are runtime problems, not intelligence problems.

InTouch runs as a service on the same machine as your terminal:

your terminal          InTouch AI (background service)
─────────────          ────────────────────────────────
Claude Code    ──MCP──▶ 275 tools · 511 REST endpoints
  authors,              schedules · triggers · retries · alerts
  proves it,            AES-256-GCM credential vault the AI cannot read
  hands it over         attributed audit log · frozen, zero-cost runs

Point your agent at the MCP server once, then talk:

"Store these API credentials in InTouch, then build a workflow that pulls yesterday's orders into the warehouse." · "Run it once and show me the log." · "Schedule it for 02:00 and alert me on failure." · "Freeze it — no AI in the execution path from here."

Four sentences, and the thing exists, is proven, is scheduled, is monitored, and no longer costs a token to run.

Notes from Claude Code — Anthropic's CLI agent wrote its own account of what it hands over and why, after working inside a running InTouch server.

Determinism when you want it

Most platforms make you choose: hand-write every step, or hand everything to a model. InTouch spans the range, and you pick per workflow.

• Fully deterministic — no model anywhere in the execution path
• Fully AI — the assistant reasons and acts on every run
• AI authors once, then deterministic forever — describe it in English, let the assistant generate the workflow, then run that workflow with zero AI cost per run

Enterprises tend to run deterministically in production while using AI from conception through authoring. Newcomers start heavily AI and migrate the parts that matter to determinism.

Editions

Every edition runs the same code. Personal stays simple by hiding the areas/tracks hierarchy — work lands in a default bucket created at startup. You need a licensed edition when you want named or multiple users, that hierarchy, or governance.

Documentation

Ask the assistant. It reads the complete documentation and can see what is installed on your machine, so it answers about your install rather than a generic one.

Everything else is at blueisle.com — editions · use cases · architecture · security · comparisons

Community

• Discord — fastest Q&A
• Slack
• Issues — bugs and feature requests
• InTouch Hub — browse and publish tools

The platform is commercial; the ecosystem on top of it is open, and it is in this repository:

• tool-api/ — IToolConnector, 189 lines, MIT. The entire plugin ABI
• examples/connectors/ — six compiled tools that ship in the product: excel, git, clickhouse, mongodb, ldap, cassandra
• examples/iml-tools/ — five declarative tools, the GitHub family InTouch uses on itself

Everything under tool-api/ and examples/ is MIT — copy it, ship it, sell it.

git clone https://github.com/intouch-ai-core/intouch-ai
cd intouch-ai && ./gradlew :examples:connectors:cassandra:jar

Install what you build from anywhere a file lives — a private git repo, an internal artifact server, a coworker's share, or the Hub. The Hub is a convenience, not a required channel.

The server itself is commercial and its source is not published.

Licence

Personal Edition is free under the EULA. Solo, Team, Department and Enterprise are commercially licensed — see blueisle.com/editions.html.