Case study · AI agents
Altrun: an AI assistant that turns messages into actions
Altrun watches the channels people actually use (Telegram, email, Instagram DMs and the browser) and turns what comes in into scheduled, personalised actions. It sorts, summarises, researches, drafts and schedules, and it learns to write the way its user does.
- Client
- Our own product
- Year
- 2026
- Our role
- Product, design and engineering
- Stack
- LLM tool calling, Agentic workflows, Telegram Bot API
The problem
What needed fixing
People who run small businesses lose hours a week triaging messages across four apps. Chat assistants answer when asked. They don't act on what's arriving, and the drafts they write don't sound like the person sending them.
The approach
How we built it
- 01
Each step is its own tool
Each capability is a tool the agent calls with structured input and output: classify, summarise, research, draft, schedule. Every step can be logged, retried and tested on its own.
- 02
One inbox shape
Messages from every channel are normalised into a single event before the agent sees them. Adding a channel means writing one adapter.
- 03
Sounds like you
Drafts are styled from the user's own past writing, so a reply reads like it came from them.
The outcome
Where it landed
In public beta and onboarding early users. It's also our test bed: patterns proven in Altrun go into client agents.
- Channels in one agent
- 4
- Channels in one agent
- Typed tools
- 5
- Typed tools
- Status, 2026
- Beta
- Status, 2026
LLM tool calling · Agentic workflows · Telegram Bot API · Email · Instagram DMs · Browser extension
Services used
We can do the same for you
- AI agentsAgents that read what comes in, decide what to do and do it, with a person in the loop where it matters.
- AI workflow automationHours of copy, paste and chase turned into flows that run on their own, with AI steps where judgement is needed.
- Chatbots & voice agentsChat and voice agents on WhatsApp, web and phone that answer, book and hand over to a person at the right moment.
Start here
Got a workflow that eats your team's week?
Tell us about it. Hussain reads every message himself and replies within a working day, usually with a question or two.