💾
Backups & data movement
Scheduled DB dumps, S3/FTP/SSH transfers, database-to-database streaming. The stuff that's a one-liner until it silently stops working.
Real automation. AI-native. Yours to drive.
InTouch AI is a full automation platform, AI-native to the core — 60+ tools, 9 LLM providers, a built-in RAG pipeline, agentic Monitors, database-to-database streaming, and a 367-tool MCP server. Build a job by describing it in plain English, write it as jobs-as-code YAML, or drive every endpoint from your editor — your call, per job. It runs on your own machine, behind your own NAT, free for personal use, no limits.
AI-native
60+ tools, 9 LLM providers
RAG + agentic Monitors
Jobs-as-code (YAML)
REST API + MCP
Self-hosted & free
InTouch AI is a general-purpose automation platform. This page is the developer's lens on it: what it replaces, what it gives you that a pile of scripts never will, and how it talks to your stack — REST, MCP, YAML, the works. It runs on your laptop today and scales to a server when you're ready.
Look What's In The Box
This is not a small tool.
Years of engineering in one self-contained server. Every number here is something you can put to work on day one — on your own machine, for free.
60+
built-in tools
9
LLM providers
8
notification channels
413
REST endpoints
367
MCP tools
14+
databases
5,000+
installable skills
Databases
14+ engines — MySQL, Oracle, PostgreSQL, SQL Server, DB2, Cloud Spanner, and more. Five SQL operation types, database-to-database streaming, batch imports up to 1000× faster.
Cloud & transfer
S3, FTP/SFTP, SSH, HTTP, Docker. Move data and drive remote systems with credentials that never leave the vault.
AI
17 native AI task types across 9 providers, a built-in RAG pipeline, and agentic Monitors. Local-and-free via Ollama, or any frontier model.
Messaging
8 outbound channels — Email, Slack, Discord, Telegram, SMS, WhatsApp, Teams, LINE. Same vault, same audit, every channel.
Enterprise apps
Purpose-built tools for Essbase, TM1, and JDE — the integrations general-purpose tools make you write from scratch.
DataFrames & transforms
In-flight data transformation with automatic dependency resolution — extract, reshape, join, and hand off to the next tool in the job.
…plus a whole Hub of ready-to-install skills, job files, and Monitors to start from instead of a blank page.
Built on the latest, both ends
A modern web UI
A clean, fast React front end — jobs, schedules, credentials, Monitors, and the AI assistant, all visual, all in the browser. Served same-origin by the server itself, no separate UI process.
A modern server
Kotlin on Micronaut 4 with compile-time DI (KSP), JVM 17, Jetty. One self-contained fat JAR, sub-second startup, ~2 GB RAM — laptop to data center, same binary.
An encrypted vault
AES-256 credential vault at the center of it all. Secrets referenced by name, never exposed to tool authors, never in plaintext on disk.
Put It To Work
The work developers point it at first
A sampler — from quick operational wins to real data and integration workloads.
📡
Monitors & watchers
SSL/cert expiry, DNS changes, endpoint health, website/competitor diffs. Fire an alert (or an AI summary) the moment something moves.
📜
Log scans & anomaly digests
Sweep logs on a schedule, let an AI step flag the anomalies, get a digest on your channel instead of grepping at 2am.
🚀
Repo & release digests
GitHub/GitLab activity rollups, dependency/PR digests, release-note drafts. Summarized, scheduled, delivered.
📊
Scheduled reports & pipelines
SQL extract → DataFrame transform → AI narrative → deliver. Multi-step jobs with real error handling, not a chain of &&.
🚨
Incident enrichment
On alert, pull recent deploys, related tickets, and past similar incidents into one structured summary — before the pager even finishes buzzing.
Drive It However You Think
Talk to it. Write it. Or program it.
The same platform, three front doors — pick per job, not once forever. Everything the UI does, the API does; everything you build is code you can version.
Talk to it
Describe the job in plain English to the AI assistant — it reads your server's state, builds the job, and runs it, with tool-use access to all 60+ tools. The fastest path from idea to running.
Write it as code
Jobs-as-code and tools-as-code in YAML you can diff, review, and commit. Convert any of 5,000+ OpenClaw skills into a deterministic InTouch job while you're at it.
Program it
413 REST endpoints (OpenAPI 3.0, full UI parity, Swagger built in) and a 367-tool MCP server to drive the entire platform from Claude Code or any MCP client.
One Toolbox, The Whole Spectrum
Learn the primitives once. Aim them at anything.
Tools, AI steps, schedules, triggers, the API — they're building blocks, and they recombine. The same handful of pieces spans a five-minute personal hack and a production data pipeline. That's the part you'll appreciate on day two.
A five-minute hack
“Ping me on Telegram when this endpoint starts 404ing.” One trigger, one tool. Done before your coffee's cold.
Ops glue
Nightly DB dump → S3 → notify on failure. A schedule and two tools, with a vault and an audit log you didn't have to build.
A real ETL pipeline
SQL extract → DataFrame transform → AI narrative → deliver to Slack. Four tools, one job, real error handling — not a chain of && and a prayer.
Integration plumbing
Inbound webhook → transform → fan out to three systems. Triggers in, HTTP tools out, every hop logged.
An agentic loop
A Monitor watches a condition, an AI step makes the call, an action fires — under your credentials, your audit. AI exactly where judgment belongs, and nowhere else.
A reusable tool of your own
Author a YAML tool or drop in a compiled plugin once; call it from every job after. The thing you built for one problem becomes the thing you reach for on ten.
Same primitives — personal to production, one-off to scheduled, deterministic to AI-driven. You learn them once and the whole spectrum of use cases opens up. Browse all 60+ tools →
As Much AI As You Want — Or None
Prototype with AI, then lock it down
Let the assistant improvise a one-off. Place an AI step exactly where you want judgment and keep the rest deterministic. Or, once a workflow is proven, convert it to a job that runs with no AI in the path — identical every time, zero token cost per run. You decide where the AI is, and where it isn't.
Straight Talk
What InTouch AI is not
So you don't download it expecting the wrong thing.
- Not a CI/CD server. Jenkins, GitHub Actions, GitLab CI own build-test-deploy. InTouch AI is for the recurring operational and integration work around your apps — and it'll happily call your pipeline.
- Not a container orchestrator. It's not Kubernetes and doesn't want to be. It runs jobs, not your cluster.
- Not an APM / observability stack. It can watch endpoints and digest logs, but it's not Datadog or Grafana. It complements them — it acts on what they see.
Sold? It installs in about five minutes.
Free for personal use, runs on your own machine, your data and your prompts never leave your network. Download the zip, or follow the step-by-step.
Blue Isle Software · Miami, FL · 954-353-BLUE