Chat in your browser.
Talk to your agent from the KilatBot dashboard. Send a photo, a file or a voice note. Replies stream in and show the tools it uses. Steer it, queue a follow-up or stop it mid-reply.
KilatBot
A private agent with memory, scheduled tasks and connected tools. Chat with it in your browser or on Telegram, give it files and let it work on a schedule.
01 / Your agent
It remembers your context, works on a schedule and uses the tools you allow. You stay in charge.
Talk to your agent from the KilatBot dashboard. Send a photo, a file or a voice note. Replies stream in and show the tools it uses. Steer it, queue a follow-up or stop it mid-reply.
Your KilatBot opens a real browser when a task needs one. Watch it beside the chat, take control for a sign-in or a payment, then hand it back. How it works →
Before a risky command runs, you see it and choose Allow or Deny. Low-risk commands run without asking.
Your agent connects to KilatBot over MCP: it can publish websites and check its own status. Add KilatBot to Claude or any other MCP client too; purchases stay in the dashboard.
Pair your Telegram bot and choose who may talk to it, let your agent remember what matters, and set tasks that run on a schedule. All from your agent's pages.
Agents start on Kilat AI, with usage included in your plan within a 5-hour and a weekly limit. Add your own provider key any time and pay your provider directly.
02 / Templates
Each one is a starting point. Change its persona, skills, schedules and tools any time.
Weekly content plans and caption drafts; you post.
Daily reports from your CSV files; no guessed numbers.
Meeting notes, weekly reviews, drafts and plans.
Builds and publishes sites, runs scripts in a sandbox.
03 / What it works with
Turn conversations into useful outputs. Publish a page, prepare a report and connect tools from the same workspace.
Ask your agent for a landing page or a menu and it builds and publishes it for you. Free websites: up to 10 per verified workspace, at Rp0. A Paid website takes your own domain when you're ready.
Ask your agent to research a topic, organize findings and create a report or an interactive artifact. Review the result in chat before sharing it.
Create an MCP token in your workspace and connect a compatible AI assistant. Check agent status, publish outputs and operate connected tools within token permissions. Purchases stay in the dashboard. Read the MCP guide →
MCP exposes persona choices and stored service selections, not verified agent readiness. It supports status, start, stop, restart and failed-deployment retry. New purchases and configuration stay in the dashboard; approve disruptive actions explicitly.
Built to work for you
Create your workspace, create your agent and give it something to do.
Get started