Using the Letta Telegram Bot

Connect your Letta agents to Telegram for conversations anywhere

The Letta Telegram Bot (@letta_ai_bot) lets you chat with your Letta agents directly through Telegram, bringing persistent memory and intelligent conversations to your favorite messaging app.

Communications with Letta agents using the Telegram bot will modify the agent’s state everywhere that Letta agents are available — for example, you will see the Telegram messages appear in Letta’s ADE.

New Features:

  • Memory Management - View and inspect your agent’s memory blocks with /blocks and /block
  • Agent Templates - Quick-start with preconfigured agents using /template
  • Reasoning Control - Toggle agent thought visibility with /reasoning enable/disable
  • Proactive Notifications - Allow agents to send you messages with /telegram-notify
  • Enhanced Navigation - Comprehensive help system and quick reference commands

Getting Started

Step 1: Find the Bot

Open Telegram and search for @letta_ai_bot or visit t.me/letta_ai_bot directly.

Step 2: Start Your First Conversation

Send /start to the bot. You’ll receive an interactive welcome with buttons to guide you through setup.

Step 3: Connect Your Letta Account

You’ll need a Letta API key to use the bot:

  1. Get Your API Key:

    • Go to app.letta.com
    • Sign in or create a free account
    • Navigate to Settings → API Keys
    • Create a new API key and copy it
  2. Login to the Bot:

    /login sk-your-api-key-here

    The bot will immediately delete your message for security and confirm successful authentication.

Step 4: Select an Agent

View your available agents:

/agents

Select an agent to chat with:

/agent agent-id-here

Step 5: Start Chatting

Once you’ve selected an agent, just send any message to start your conversation.

Essential Commands - Complete Reference

This comprehensive guide covers all available Telegram bot commands, organized by category for easy navigation. Commands marked with (star) are essential for getting started.

Quick Navigation:

Authentication Commands

/start (star) - Welcome and Setup Guide

Opens the interactive setup wizard for new and returning users.

Behavior:

  • For new users: Shows welcome message with link to get API key
  • For authenticated users: Shows welcome back message

Example Output (New User):

(hey sarah, welcome to letta)
looks like you're new here. want help getting started?
[sure] [i know what i'm doing]

Example Output (Returning User with Agent):

(welcome back sarah. you're chatting with Ion)
[switch agent] [view tools] [just chat]

Example Output (Returning User, No Agent):

(welcome back sarah. want to pick an agent?)
[show my agents]
[create Ion]
[maybe later]

/login <api_key> (star) - Authenticate with Letta

Connects your Letta account to the Telegram bot. Your API key is immediately deleted from chat history for security.

What it does:

  1. Validates your API key with Letta’s servers
  2. Encrypts and stores your credentials securely
  3. Deletes the message containing your API key
  4. Confirms successful authentication

Example Command:

/login sk-1234567890abcdef...

Expected Output (Success, Has Agents):

(all set. welcome sarah)
want to pick an agent?
[show my agents]
[create Ion]
[maybe later]

Expected Output (Success, No Agents):

(all set. welcome sarah)
looks like you need an agent. want to create one?
[create Ion]

Expected Output (Invalid Key):

Authentication failed
The API key appears to be invalid. Please check:
1. You copied the complete key
2. The key hasn't expired
3. You're using the correct key
Get your API key at: https://app.letta.com

/logout - Remove Stored Credentials

Completely removes your stored API key and preferences from the bot.

What it does:

  1. Deletes your encrypted credentials
  2. Clears your agent selections
  3. Removes all stored preferences

Expected Output:

Logout successful
Your credentials have been removed.
Use /login <api_key> to authenticate again.

/status - Check Authentication Status

Shows your current authentication status and configuration.

Expected Output (Authenticated):

Authentication Status
Authenticated: Yes
API URL: https://api.letta.com
Last Updated: 2024-01-15 10:30:00
You're ready to chat with your agents!

Expected Output (Not Authenticated):

Not authenticated
Please use /login <api_key> to authenticate.
Get your API key at: https://app.letta.com

Agent Management Commands

/agents (star) - List Available Agents

Shows all agents in your current project with their IDs.

Expected Output:

(your agents)
currently using: research assistant
available (3):
• research assistant
`agent-abc123def456`
• personal helper
`agent-xyz789ghi012`
• code reviewer
`agent-mno345pqr678`
[use personal helper]
[use code reviewer]
[← prev] [next →]

/agent - Show Current Agent

Displays information about your currently selected agent.

Expected Output (Agent Selected):

**Current Agent**
Name: Research Assistant
ID: `agent-abc123def456`
Model: gpt-4
Created: 2024-01-10
Memory Blocks:
• human (2000 chars)
• persona (2000 chars)
• archival_memory (10000 chars)
Use /ade to edit this agent in the web interface.

Expected Output (No Agent):

No agent selected
Use /agents to see available agents
Use /agent <agent_id> to select one

/agent <id> (star) - Select an Agent

Switches to a specific agent for your conversation.

Example Command:

/agent agent-xyz789ghi012

Expected Output (Success):

Agent selected successfully!
Now chatting with: Personal Helper
This agent will be used for all messages in this chat.

Expected Output (Not Found):

Agent not found: agent-invalid123
This agent doesn't exist or you don't have access to it.
Use /agents to see your available agents.

/ade - Agent Development Environment

Provides a direct link to edit your current agent in Letta’s web interface.

Expected Output:

Agent Development Environment
Agent: Research Assistant
Link: https://app.letta.com/agents/agent-abc123def456
Click the link to edit your agent's:
• System prompts and personality
• Memory blocks
• Tool configuration
• Model settings

Project Management Commands

/projects - List All Projects

Shows all projects in your Letta workspace.

Expected Output:

(projects)
currently using: main workspace
`project-123abc`
available (3):
• main workspace (3 agents)
`project-123abc`
• development (5 agents)
`project-456def`
• personal (2 agents)
`project-789ghi`
[use development]
[use personal]
[← prev] [next →]

/project - Show Current Project

Displays information about your current project.

Expected Output:

**Current Project**
Name: Main Workspace
ID: `project-123abc`
Agents: 3
Created: 2024-01-01
Use /projects to see all projects
Use /project <id> to switch projects

/project <id> - Switch Project

Changes to a different project, which changes the available agents.

Example Command:

/project project-456def

Expected Output:

Project switched successfully!
Now using: Development
5 agents available
Note: Your previous agent selection has been cleared.
Use /agents to see agents in this project.

Tool Management Commands

/tool - Interactive Tool Management

Opens an interactive menu for managing tools attached to your agent.

Expected Output:

(tools for research assistant)
attached tools (3):
• web_search
• calculator
• send_telegram_message
[attach new tools]
[detach tools]
[back to main menu]

Attach Menu (with pagination): Selecting “attach new tools” shows available tools with navigation:

(available tools - page 1/3)
[web_search]
[calculator]
[code_interpreter]
[wikipedia_search]
[weather_api]
[news_api]
[translator]
[file_processor]
[← prev] [next →] [back]

Detach Menu: Selecting “detach tools” shows all attached tools:

(remove tools - 3 attached)
[web_search]
[calculator]
[send_telegram_message]
[back to tools]

/tool attach <name> - Direct Tool Attachment

Directly attaches a tool without using the interactive menu.

Example Command:

/tool attach web_search

Expected Output (Success):

(tool attached)
web_search is now available to your agent
search the internet for current information and recent data

Expected Output (Already Attached):

(already attached)
web_search is already available to this agent

Expected Output (Not Found):

(tool not found)
invalid_tool doesn't exist. use /tool to see available options

/tool detach <name> - Direct Tool Removal

Directly removes a tool without using the interactive menu.

Example Command:

/tool detach web_search

Expected Output:

(tool detached)
web_search has been removed from your agent

Shortcut Commands

/shortcut - List Saved Shortcuts

Shows all your quick-access shortcuts.

Expected Output:

**Your Shortcuts**
work → Research Assistant
`agent-abc123def456`
personal → Personal Helper
`agent-xyz789ghi012`
code → Code Reviewer
`agent-mno345pqr678`
**Usage:**
• `/switch <name>` to quickly switch agents
• `/shortcut <name> <agent_id>` to create new
• `/shortcut delete <name>` to remove

/shortcut <name> <agent_id> - Create Shortcut

Creates a quick-access shortcut for an agent.

Example Command:

/shortcut research agent-abc123def456

Expected Output:

**Shortcut Created**
`research` → Research Assistant
Now you can use `/switch research` to quickly select this agent.

/switch <name> - Quick Switch

Instantly switches to an agent using its shortcut.

Example Command:

/switch work

Expected Output:

**Switched via shortcut**
Now chatting with: Research Assistant
(shortcut: work)

/shortcut delete <name> - Remove Shortcut

Deletes a saved shortcut.

Example Command:

/shortcut delete old_agent

Expected Output:

**Shortcut Deleted**
Shortcut `old_agent` has been removed.

Help Commands

/help - Command Reference

Shows a concise list of all available commands.

Expected Output:

Commands:
/start - Setup guide
/login <api_key> - Authenticate
/logout - Remove credentials
/status - Check authentication
/project - Show/switch project
/projects - List projects
/agent - Show/switch agent
/agents - List agents
/make-default-agent - Create default agent
/template - List and create agent templates
/ade - Get agent web link
/tool - Manage tools
/telegram-notify - Enable proactive notifications
/shortcut - Manage shortcuts
/switch <name> - Quick switch
/blocks - List memory blocks
/block <label> - View memory block
/reasoning enable|disable - Show/hide reasoning messages
/clear-preferences - Reset preferences
/refresh - Update cached agent info
/help - Show commands

Memory Management Commands

/blocks - List Memory Blocks

Shows all memory blocks for your current agent with their labels and character usage.

Expected Output:

**Memory Blocks for Research Assistant**
**persona** (1,245/2,000 chars)
Core personality and role definition
**human** (891/2,000 chars)
Information about the human user
**tool_use_guidelines** (542/2,000 chars)
Guidelines for using available tools
**communication_guidelines** (423/2,000 chars)
How to communicate effectively
**procedures** (667/2,000 chars)
Standard operating procedures
**scratchpad** (234/2,000 chars)
Temporary storage for ideas and notes
Use `/block <label>` to view a specific block

/block <label> - View Memory Block

Displays the contents of a specific memory block.

Example Command:

/block persona

Expected Output:

**Memory Block: persona**
Description: Core personality and role definition
Content:
I am a thorough research assistant who helps find, analyze, and synthesize information. I'm curious and detail-oriented, focusing on providing comprehensive answers while remaining engaging and helpful. I adapt my communication style to match the user's needs and preferences.

Expected Output (Block Not Found):

(error: block 'invalid_block' not found - use /blocks to see available blocks)

Template Management Commands

/template - List Agent Templates

Shows available agent templates for quick agent creation.

Expected Output:

**Available Templates**
• **Ion** - adaptive AI with infinite memory that develops theories about you
Use: `/template <name>` or click below
[Ion]

/template <name> - Create Agent from Template

Creates a new agent using a predefined template.

Example Command:

/template ion

Expected Output (Success):

(Ion is ready - we've asked Ion to greet you, please wait)
Hi! I'm Ion, your new AI assistant. Unlike other AIs, I remember everything from our conversations and develop theories about how you think over time.
Here are some ways to get started:
• Share a link or article you'd like to discuss
• Ask me to research something you're curious about
• Tell me about yourself - interests, work, what excites you
• Give me a problem to think about with you
• Or just start talking - I learn from everything we discuss
I'll remember this conversation forever and build on it next time we chat.

/make-default-agent - Create Default Agent

Creates a simple Ion agent when you have no agents configured.

Expected Output:

(creating agent Ion)
(Ion is ready)
(**Ion** says)
Hello! I'm Ion, working with you now. I'm a stateful agent with persistent memory that adapts to your communication style naturally.
What would you like to work on together?

Notification Management Commands

/telegram-notify - Manage Proactive Notifications

Controls whether your agent can send proactive notifications via Telegram.

Usage Options:

  • /telegram-notify or /telegram-notify status - Check current status
  • /telegram-notify enable - Allow proactive notifications
  • /telegram-notify disable - Disable proactive notifications

Expected Output (Status):

**Telegram Notifications**
Agent: Research Assistant
Status: Enabled
Your agent can send proactive notifications when:
• Important events occur
• Scheduled reminders trigger
• Research tasks complete
Use `/telegram-notify disable` to turn off

Expected Output (Enable):

**Notifications Enabled**
Research Assistant can now send proactive Telegram messages.
The 'send_telegram_message' tool has been attached to your agent.

Expected Output (Disable):

**Notifications Disabled**
Research Assistant will no longer send proactive Telegram messages.
The 'send_telegram_message' tool has been removed from your agent.

Preference & Settings Commands

/reasoning enable|disable - Control Reasoning Messages

Controls whether you see your agent’s internal reasoning/thought processes.

Example Commands:

/reasoning enable
/reasoning disable

Expected Output (Enable):

Reasoning messages enabled

Expected Output (Disable):

Reasoning messages disabled

What are reasoning messages? When enabled, you’ll see messages like:

(**Research Assistant** thought)
> I need to search for recent information about this topic since my knowledge might be outdated. Let me use the web search tool to find current data.

When disabled, you’ll only see the agent’s final responses, not their internal thinking process.

/refresh - Update Cached Agent Info

Updates the bot’s cached information about your current agent (name, etc).

Expected Output:

**Agent Info Refreshed**
Updated cached info for: Advanced Research Assistant
(was: Research Assistant)
Agent name and details are now current.

Expected Output (No Changes):

**Agent Info Current**
Research Assistant - no updates needed
Cached information matches current agent state.

/clear-preferences - Reset User Preferences

Clears all your stored preferences and settings (debug command).

Expected Output:

(preferences cleared)

Expected Output (No Preferences):

(no preferences found)

Understanding Your Agents

What Are Letta Agents?

Letta agents are AI assistants with persistent memory. Unlike regular chatbots, they:

  • Remember Everything: Your conversation history is preserved across sessions
  • Maintain Context: They understand references to previous discussions
  • Use Tools: Can search the web, run calculations, and more
  • Have Personalities: Each agent can have unique traits and knowledge

Managing Multiple Agents

You can create different agents for different purposes:

  • Personal Assistant: General help and task management
  • Research Agent: Specialized in finding and analyzing information
  • Creative Partner: Focused on brainstorming and creative work
  • Study Buddy: Helps with learning and retention

Switch between agents anytime with /agent <id> or use shortcuts for quick access.

Agent Templates

When you first connect your Letta account or need to create new agents, the bot offers quick-start templates to get you up and running immediately.

Available Templates

Ion: Advanced AI assistant with sophisticated memory architecture

  • Features 6 specialized memory blocks for comprehensive understanding
  • Adapts to your communication style and remembers preferences over time
  • Professional yet conversational, matching your energy level
  • Includes memory management tools for continuous learning
  • Perfect for users who want an intelligent, adaptive assistant

Using Templates

Templates appear as buttons when you first login or when you have no agents:

(all set. welcome sarah)
looks like you need an agent. want to create one?
[create Ion]

Simply tap the Ion button to instantly create and configure a new agent with advanced memory capabilities and adaptive personality.

Ion’s Advanced Memory System

Ion features a sophisticated 6-block memory architecture that enables deeper, more contextual conversations:

Core Memory Blocks:

  1. Persona: Ion’s adaptive personality that evolves to complement your communication style
  2. User Profile: Dynamic profile that learns your preferences, interests, and patterns
  3. Memory Directives: Guidelines for how Ion manages and updates its memory
  4. Interaction Patterns: Tracks your communication preferences and response styles
  5. Knowledge Graph: Builds semantic connections between topics you discuss
  6. Temporal Context: Understands time-based patterns and routines in your life

Key Features:

  • Proactive memory management using specialized tools
  • Natural learning through conversation without interrogation
  • Cross-conversation context retention
  • Adaptive personality that matches your energy level
  • Professional communication without excessive enthusiasm

Ion is designed to be your intelligent, adaptive assistant that gets better at helping you over time.

Advanced Features

Interactive Navigation

The bot uses interactive buttons throughout for easier navigation:

  • Command responses include relevant action buttons
  • Menu systems for tools, agents, and projects
  • Pagination controls for long lists
  • Quick actions accessible without typing commands

Example interactive flow:

/start → [sure] → /login key → [create Ion] → ready to chat

Using Tools

Your agents can use various tools to help you:

Web Search: Agents can search the internet for current information

User: What's the weather like in San Francisco today?
Agent: Let me search for current weather information... [searches web]

View Available Tools:

/tool

Attach New Tools:

/tool attach web_search

The bot automatically handles large lists with pagination:

Agent Lists: Shows 5 agents per page with next/previous buttons Tool Attachment: Shows 8 tools per page when attaching Tool Detachment: Shows all attached tools (no pagination) Projects: Paginated display for multiple projects

Creating Shortcuts

Save time with shortcuts for frequently used agents:

Create a Shortcut:

/shortcut work agent-abc123
/shortcut personal agent-xyz789

Use Shortcuts:

/switch work
/switch personal

Working with Projects

Organize your agents by project:

List Projects:

/projects

Switch Projects:

/project project-id-here

Privacy and Security

Your Data Is Protected

  • API Key Encryption: Your API key is encrypted on the bot server with a unique key
  • Isolated Access: You can only see your own agents and data
  • Secure Communication: All messages are transmitted securely
  • Letta Cloud Storage: Your conversations and agent data are stored on Letta Cloud and may be visible there

Best Practices

  1. Protect Your API Key: Never share it with others
  2. Use /logout: Always logout on shared devices
  3. Regular Checks: Use /status to verify your session
  4. Report Issues: Contact support if you notice anything unusual

Troubleshooting

Bot Not Responding

If the bot doesn’t respond:

  1. Check your internet connection
  2. Verify you’re logged in with /status
  3. Try selecting an agent again with /agent <id>
  4. Send /help to reset the conversation

Authentication Issues

If you can’t login:

  1. Verify your API key at app.letta.com
  2. Make sure you’re copying the entire key
  3. Try generating a new API key
  4. Use /logout first if you’re having issues

Agent Not Working

If your agent isn’t responding properly:

  1. Check that an agent is selected with /agent
  2. Verify the agent exists with /agents
  3. Try switching to another agent and back
  4. Visit the ADE with /ade to check agent status

Message Errors

If you see error messages:

  • “Authentication required”: Use /login with your API key
  • “No agent selected”: Choose an agent with /agent <id>
  • “Agent not found”: The agent may have been deleted; use /agents to see available ones
  • “Timeout”: Complex requests may take time; try again or simplify your request

Common Use Cases

Daily Planning

You: Good morning! What's on my schedule today?
Agent: Good morning! Based on our previous discussions, you have:
1. Team meeting at 10 AM about the Q4 roadmap
2. Lunch with Sarah to discuss the marketing campaign
3. Code review for the new feature at 3 PM

Research Assistant

You: I need recent information about sustainable packaging trends
Agent: Let me search for the latest developments... [searches web]
Here are the key trends in sustainable packaging for 2024...

Learning Partner

You: Can you quiz me on the Python concepts we studied yesterday?
Agent: Of course! Let's start with the list comprehensions we covered...

Creative Brainstorming

You: Help me come up with names for my new coffee shop
Agent: Based on the vibe you described (cozy, bookish, vintage), here are some suggestions...

Practical Examples and Edge Cases

Setting Up Multiple Agents for Different Purposes

Scenario: You want different agents for work and personal use.

# First, create shortcuts for easy switching
/shortcut work agent-abc123
/shortcut personal agent-xyz789
# During work hours
/switch work
"Can you review the quarterly report I mentioned yesterday?"
# After work
/switch personal
"What ingredients did we decide on for the dinner party?"

Handling Authentication Issues

Edge Case: Your API key expires or becomes invalid.

You: Hello, are you there?
Bot: Authentication error
Your stored credentials appear to be invalid.
Please try /logout then /login <new_api_key>
# Solution:
/logout
/login sk-new-api-key-here

Managing Tools for Specific Tasks

Scenario: You need web search for a research project.

# Check current tools interactively
/tool
# Select "attach new tools" button
# Navigate to web_search and select it
# Or use direct command:
/tool attach web_search
You: What are the latest developments in quantum computing?
Agent: Let me search for recent information... [uses web_search tool]

Working with Large Agent Lists

Edge Case: You have many agents and can’t remember IDs.

# Use shortcuts for frequently used agents
/shortcut main agent-abc123def456
# Search projects if agents are organized by project
/projects research
/project project-research-001
/agents
# Use descriptive shortcut names
/shortcut research-quantum agent-qnt789
/shortcut research-climate agent-clm456

Recovering from Errors

Scenario: Bot stops responding mid-conversation.

# First, check your status
/status
# If authenticated, check agent selection
/agent
# If no agent selected, reselect
/agents
/agent agent-abc123
# If issues persist, refresh credentials
/logout
/login sk-your-api-key

Switching Context Quickly

Scenario: Multiple ongoing projects with different agents.

# Morning standup prep
/switch work
"Summarize yesterday's progress on the API integration"
# Client meeting
/switch client-project
"What were the three main concerns from our last meeting?"
# Personal reminder
/switch personal
"What time did I say I'd pick up the kids?"

Tool Conflicts and Resolution

Edge Case: Multiple tools with similar names.

/tool attach search
Bot: (ambiguous tool name)
multiple tools match 'search':
• web_search - search the internet
• doc_search - search documents
• code_search - search codebases
try being more specific
# Solution:
/tool attach web_search

Memory Continuity Across Sessions

Example: Continuing a complex discussion.

# Monday
You: Let's plan the product launch for next month
Agent: I'll help you plan the launch. Let's start with...
# Wednesday
You: Where did we leave off with the launch plan?
Agent: On Monday, we outlined the timeline and identified three key milestones...

Handling Network Issues

Edge Case: Slow or interrupted connections.

You: [Sends message]
[No response for 30 seconds]
# Bot will show typing indicator if processing
# If timeout occurs:
Bot: (please wait)
that took longer than expected. the query might be complex - try breaking it into smaller parts

Project and Agent Coordination

Scenario: Organizing agents across multiple projects.

# View all projects
/projects
# Switch to development project
/project project-dev-123
# See dev-specific agents
/agents
# Create shortcuts for cross-project access
/shortcut dev-backend agent-backend-456
/shortcut dev-frontend agent-frontend-789

Getting Help

Quick Help

  • Send /help for a command reference
  • Send /start for the setup guide

Documentation

Support

Frequently Asked Questions

Q: Is the bot free to use? A: The bot itself is free, but you need a Letta account. Letta offers free tiers with usage limits.

Q: Can I use multiple agents in the same chat? A: Yes! Switch between agents anytime with /agent <id> or shortcuts.

Q: Will my agent remember conversations from other platforms? A: Yes, if you’re using the same agent, it maintains memory across all platforms.

Q: How long are conversations stored? A: Your agent’s memory persists indefinitely as part of your Letta account.

Q: Can I share my agent with others? A: Each user needs their own Letta account and API key to use the bot.

Q: What happens if I delete an agent? A: The agent and all its memories are permanently removed from your Letta account.

Q: Can I use the bot in group chats? A: The bot is designed for individual use. Each user needs their own authentication.

Quick Reference Card

Essential Commands:
/start - Setup guide
/login <key> - Connect account
/agents - List agents
/agent <id> - Select agent
/template <name> - Create from template
/help - Command list
Quick Actions:
/switch <name> - Use shortcut
/ade - Edit agent
/status - Check status
/logout - Disconnect
Management:
/projects - List projects
/tool - Manage tools
/shortcut - Manage shortcuts
/blocks - View memory blocks
/block <label> - View specific block
Settings:
/reasoning enable|disable - Toggle thoughts
/telegram-notify enable - Proactive messages
/refresh - Update agent info
/clear-preferences - Reset settings

Start chatting with your intelligent, memory-equipped AI agents today at t.me/letta_ai_bot!