Web platform · cloud and self-hosted
Automation you can see
FalangFlow is a visual automation builder: integrations between services, bots and AI scenarios. Every process is a diagram that reads top to bottom and runs as a durable Temporal workflow — with retries, timers and recovery after a crash. Describe the task in words, and the agent builds the process for you.
How it works
A Telegram message, a webhook, a schedule, a Bitrix24 event or a confirmed payment.
Yourself, from blocks and integrations — or ask the AI agent in plain language.
Runs with the current step highlighted, breakpoints and variable values.
One click to production — with version history and rollback to any earlier version.
Integrations
From Telegram and OpenAI to databases, payments and the major Russian business services. Every integration has its own encrypted credentials for test and production.
A catalog of ready-made integrations built on the open ActivePieces library — running in an isolated service, with one-click OAuth2.
FalangFlow
The whole automation lifecycle — from trigger to production — on one platform.
Triggers
A process starts on a bot message or a /start command, an incoming webhook, a schedule, a Bitrix24 event or a confirmed YooKassa payment.
TypeScript inside
Block fields are real TypeScript with strict type checking. The editor knows every variable at that point in the diagram, suggests fields and underlines typos before anything runs.
AI in the process
Calling a model is just a block: write a text, pick a branch by what a message means, draw an image or transcribe a voice note. OpenAI-compatible models, GigaChat and YandexGPT.
AI agent
Tell the chat what you want automated. The agent asks clarifying questions, creates triggers and functions, wires up integrations and builds the diagrams in front of you.
Magic node
Click a line and write what the step should do. The agent builds real blocks underneath, while the canvas keeps your text — the diagram reads like a description of the task.
Human in the loop
A “human task” block pauses the process until you decide — approve a discount, confirm a refund. The process will wait an hour, a day or a week, and won’t forget a thing.
Debugging
Run a function on the test stand and the current block lights up on the canvas. Set a breakpoint: execution pauses and the panel shows every variable in scope.
Version history
A project keeps its history like a Google Doc. Two versions open side by side, with added, removed and changed blocks highlighted right on the canvas.
MCP
Connect Claude Code or another agent with a personal access token. It reads the project, edits diagrams, builds, runs and publishes — while the diagram it’s editing is locked in your browser.
Use cases
The bot asks which bouquet to make, AI writes the greeting card, the customer pays by link, a deal is created in the CRM and the florist gets a task.
Every 15 minutes, stock and orders are reconciled between marketplaces and the warehouse — no more manual spreadsheet exports.
A model triages each website lead: hot ones go straight to a manager for approval, the rest go to a mailing list.
Deployment
No installs, no servers: build processes right in the browser, with infrastructure and updates on us.
The same engine on your side: docker compose to start, Kubernetes for production. Your data never leaves your perimeter.
FalangFlow
Retries on failure, timers that last weeks, and resuming from the same step after a server restart.
PostgreSQL and MySQL: “Sync structure” turns tables into types for select, insert, update and delete.
Telegram photos, storage with public links, and ffmpeg: thumbnails, trimming, conversion.
Intervals and cron with a visual builder — on durable Temporal Schedules.
Any API through an HTTP request block, any incoming request through a webhook.
Users and roles, limits, platform OAuth apps and AI agent settings.
Credentials are stored separately for test and production and encrypted before saving.
Route selected integrations through a proxy in another region; everything else goes direct.
Start from a ready-made process instead of a blank page.
FalangFlow
A bot takes orders and payments, a deal lands in the CRM, a task goes to whoever fulfils it. No backend hire needed.
Stock and orders are reconciled between marketplaces and your warehouse on a schedule — no manual exports.
Temporal durability, strict TypeScript and self-hosting — without a heavyweight BPM rollout.
FalangFlow
Diagrams read top to bottom instead of being a tangle of wires, fields are strictly typed TypeScript, and every process runs as a durable Temporal workflow: it survives crashes and can wait days for an event. And it’s open source.
Yes. The engine is open source under AGPL-3.0 and starts with docker compose; every process runs in its own Kubernetes pod — exactly like in the cloud.
Temporal persists the state of every step, so after a restart the process continues from where it left off. Retries and timers work out of the box.
Any OpenAI-compatible one — OpenAI, Azure OpenAI, a local llama.cpp and more; processes can also use GigaChat and YandexGPT. The AI agent’s model is set by an administrator.
Credentials are encrypted and kept separately for the test stand and production; a project export never contains them.
Self-host FalangFlow today — the source is open. The hosted version is coming soon.
Falang
A visual IDE for algorithms and code
Build algorithms as flowcharts and get real source code in C++, Go, Rust, C# and TypeScript. Fully offline, projects are plain files.
Arduino without fighting C++ syntax
Draw setup() and loop() as diagrams, drive pins and sensors with ready-made blocks, flash the sketch to a board and debug it right on the hardware.