Session Routing: Multiple Users, One Agent
Handle multiple users with separate conversation contexts and permissions, all managed by a single OpenClaw agent.
What You Will Get
After this setup, your single OpenClaw agent will handle conversations with multiple users simultaneously, each with their own isolated context, preferences, and permission level. No user's conversation will leak into another's session.
Session routing is essential when your agent serves more than one person. Without it, the agent treats all messages as part of one conversation, which leads to confused responses and potential data leaks between users. Proper session isolation means each user gets a dedicated conversation thread with its own memory and context.
You will configure how users are identified, how sessions are created and destroyed, and how the agent switches between contexts when processing messages from different users across different channels.
Step-by-Step Setup
Configure session routing for multi-user operation.
Enable Session Routing
Navigate to your agent's Settings and toggle on Session Routing in the Multi-User section. This activates per-user session management. The agent will now create a separate conversation context for each unique user it interacts with, rather than maintaining a single shared context.
Configure User Identification
Set up how the agent identifies unique users. For most channels, the platform's native user ID is the best identifier: phone number for WhatsApp, user ID for Telegram and Discord, workspace member ID for Slack. Configure fallback identification for channels without built-in user IDs.
Set Session Lifecycle Rules
Define when sessions are created and when they expire. A new session starts automatically when an unknown user sends their first message. Configure session timeout to automatically close inactive sessions after a set period, such as 24 hours. Expired sessions free up memory and start fresh if the user returns.
Configure Context Isolation
Set the level of context isolation between sessions. Full isolation means each user's conversation history, variables, and state are completely separate. Partial isolation allows sharing read-only data like a global knowledge base while keeping conversation history private. Choose the level appropriate for your use case.
Set Per-User Permissions
Assign permission levels to individual users or user groups. Admin users might have access to agent configuration commands, while regular users can only chat. Create permission profiles and assign them during user identification. The agent checks permissions before executing any privileged action.
Enable Session Persistence
Turn on session persistence to save conversation state across agent restarts. Without persistence, sessions are lost when the agent restarts and users must start fresh. Persistent sessions store context, preferences, and conversation history in durable storage so everything survives restarts and deployments.
Test Multi-User Scenarios
Have two or more test users message the agent simultaneously from different accounts. Verify that each user gets responses based on their own conversation context. Check that one user's questions do not influence another user's responses. Review session data in the RunTheAgent dashboard to confirm isolation is working correctly.
Tips and Best Practices
Set Reasonable Session Limits
Configure a maximum number of concurrent sessions to prevent resource exhaustion. If the limit is reached, new users receive a polite message asking them to try again later.
Use Session Labels
Tag sessions with labels like active, idle, or escalated. This makes it easy to monitor and manage sessions from the dashboard, especially when handling a large number of concurrent users.
Clean Up Expired Sessions
Schedule automatic cleanup of expired sessions to free storage space. Set a retention policy for conversation history so old data does not accumulate indefinitely.
Frequently Asked Questions
Related Pages
Ready to get started?
Deploy your own OpenClaw instance in under 60 seconds. No VPS, no Docker, no SSH. Just your personal AI assistant, ready to work.
Starting at $24.50/mo. Everything included. 3-day money-back guarantee.
