How to set up a multi-threaded OpenClaw workspace in Telegram

You have one OpenClaw instance. You want it to handle multiple projects, each with its own context. Maybe a work project, a side hustle, and personal tasks. Or different clients. Or different personas for different use cases.

The problem: everything lands in one chat. Context bleeds. Your agent mixes up which project you are talking about. It is chaos.

The solution: Telegram topics.

This guide shows you how to set up a multi-threaded OpenClaw workspace using Telegram's topic feature. Each topic becomes an isolated workspace with its own conversation history. One instance, multiple contexts, zero bleed.

Credit: This setup was originally shared by @chrysb. This post expands on their configuration.

What you need

• OpenClaw installed and running
• A Telegram account
• A Telegram bot token (you probably have this already if you are using OpenClaw with Telegram)

How Telegram topics work

Telegram groups have a feature called "Topics" (also called "Forum mode"). Instead of one big chat, you get multiple threads—like Slack channels within a single group.

Each topic has its own:

• Message history
• Thread ID
• Isolated context

OpenClaw can distinguish between topics. When you message in a specific topic, OpenClaw treats it as a separate conversation. Memory stays isolated. Context stays clean.

Step 1: Create a Telegram group with topics enabled

1. Open Telegram and create a new group
2. Add your OpenClaw bot to the group
3. Go to group settings → Edit → Topics
4. Enable "Topics" (Forum mode)

Your group now supports multiple threads.

Step 2: Create topics for each workspace

Click "Create Topic" and make one for each context you want:

• Work — your day job tasks
• Side Project — that app you are building
• Research — articles, papers, learning
• Personal — life admin, reminders

Or organize by client:

• Client: Acme Corp
• Client: Startup XYZ
• Client: Freelance Gigs

Whatever makes sense for your workflow.

Step 3: Configure OpenClaw for topic awareness

OpenClaw handles topics automatically when using the Telegram channel. Each topic gets its own session, so conversation history is isolated.

In your OpenClaw config, make sure you have the Telegram channel enabled:

channels:
  telegram:
    enabled: true
    token: YOUR_BOT_TOKEN
    allowedUsers:
      - your_telegram_id

That is it for the basic setup. OpenClaw will now treat each topic as a separate workspace.

Step 4: Customize SOUL.md per topic (optional)

If you want different behavior per workspace, you can use topic-specific instructions. In your main SOUL.md, add conditional logic:

# Base Persona

You are my personal assistant.

## Context-Specific Behavior

When I message you in a topic named "Work":
- Be professional and concise
- Focus on productivity and task management
- Use formal language

When I message you in a topic named "Side Project":
- Be more creative and exploratory
- Suggest ideas freely
- Keep track of feature ideas and TODOs

When I message you in a topic named "Research":
- Summarize information clearly
- Always cite sources
- Ask clarifying questions before deep dives

OpenClaw can see the topic name in the message metadata and adjust accordingly.

Advanced: Separate skills per topic

You can also load different skills based on context. For example:

• Work topic → GitHub, Slack, Calendar skills
• Research topic → Web search, Summarizer, PDF reader skills
• Personal topic → Reminders, Notes, Weather skills

To do this, create wrapper skills that check the topic before executing, or use conditional skill loading in your SOUL.md.

Why this works better than separate bots

You could run multiple OpenClaw instances, each with its own bot. But that means:

• Multiple processes to manage
• Multiple API key configurations
• Multiple places to check for responses
• Higher resource usage

The topic approach gives you:

• One process, one bot, one config
• All conversations in one Telegram group
• Easy switching between contexts (just tap a topic)
• Shared base configuration with context-specific overrides

It is simpler and cleaner.

Real-world example

Here is how one user structures their topics:

Each topic stays focused. No context mixing. When they switch topics, they switch mental contexts too.

Tips for success

Name topics clearly. OpenClaw uses the topic name for context. "Work" is better than "Thread 1."

Pin important info. Telegram lets you pin messages in topics. Pin your project goals or key reference info.

Use topic icons. Emoji help you scan and switch faster. 🏢🚀📚🏠💰

Do not go overboard. Three to five topics is plenty for most people. Too many defeats the purpose.

Archive old topics. When a project ends, archive the topic instead of deleting. You keep the history.

Limitations

A few things to know:

• Topic awareness requires the Telegram channel (not direct messages)
• The bot needs to be a group admin to see all topic messages
• Very old Telegram clients might not support topics properly

None of these are dealbreakers for most setups.

Wrap-up

Telegram topics turn one OpenClaw instance into a multi-workspace powerhouse. Different projects, different contexts, same simple setup.

The key insight is that you do not need multiple bots or complex routing. Telegram's native topic feature does the isolation for you. OpenClaw just respects those boundaries.

Set it up in ten minutes. Thank @chrysb for sharing the idea.