ACT by Boost.space
ACT is an AI agent that builds and runs your automations. Tell it what should happen and when, in plain words. It writes the code, tests it, sets the schedule, keeps your API keys safe and switches the automation on. When something should change, you ask again.

What it's for
Work that repeats, and that nobody should have to do by hand:
- Reports on a schedule. "Every weekday at 7:00, fetch the ECB euro rates for USD, GBP and CZK."
- Messages to your team. "Every Monday at 9:00, post last week's rates to our Slack channel."
- Reacting to other apps. "When our e-shop sends an order, calculate the total with 21% VAT."
- Reacting to your data. "When a new lead is created, score it and notify me."
How you use it
- Talk to the agent. Open Chat in ACT, or talk to ACT from ChatGPT, Claude or another assistant connected over MCP.
- Check the result. Automations shows everything the agent built: the trigger, the code and every run.
- Ask for changes. "Run it at 8:00 instead" is all it takes.
Why it works
- No code to write, nothing to host. The agent writes the automation and ACT runs it.
- Works with any service that has an API. The agent writes the code that calls it; you give it the key.
- Tested before it goes live. The agent test-runs the code and fixes what fails before it switches the automation on.
- Keys stay secret. API keys go into a secure form, are stored encrypted, and never appear in the chat or the code.
- Every run isolated and recorded. Each run gets its own fresh environment, and leaves behind its result or the exact error.
Start here
- Quickstart — your first automation from one request
- Work with the agent — what to ask, and how
- API & MCP — use ACT from ChatGPT, Claude and other assistants
- ACT for AI agents — the technical reference