Core Concepts

Understanding these core concepts will help you get the most out of Lenno.

What Is an Agent?

An agent is a persistent AI-powered identity that can communicate with users across multiple channels. Each agent has its own:

• Personality – instructions that define character, tone, expertise, and behavior
• Provider – the AI model powering its responses (Claude, GPT, Mistral, etc.)
• Channels – where it communicates (Telegram, Slack, Discord, email)
• Memory – persistent context that carries across conversations
• Skills – specialized capabilities and tools
• Container – an isolated compute environment where it runs

Agents are always-on. They can respond to messages, run scheduled tasks, and interact with external services.

Providers

Lenno uses a Bring Your Own Key (BYOK) model for AI providers. You supply your own API keys, which means:

• Full cost transparency – you pay the provider directly at their published rates
• No markup – Lenno does not add any surcharge on AI usage
• Provider flexibility – switch between providers or models at any time
• Rate limit control – your keys, your limits

Supported providers include Anthropic (Claude), OpenAI (GPT), and Mistral. Each provider offers multiple models with different capabilities and price points.

See Provider Setup Guides for detailed instructions.

Channels

Channels are the communication frontends where users interact with your agent. Supported channels:

• Telegram – the most feature-rich integration with real-time streaming
• Slack – connect to your workspace channels
• Discord – run a bot in your Discord server
• Email – receive and respond to emails
• Web Chat – built-in chat interface in the Lenno dashboard

Each agent can be connected to multiple channels simultaneously. Conversations are threaded per-channel, so context stays separate.

See Channel Setup Guides for detailed instructions.

Skills

Skills give your agent specialized abilities beyond basic conversation. Skills are essentially pre-built instruction sets and tool combinations that enable specific workflows.

Examples of what skills can do:

• Search the web for current information
• Read and summarize documents
• Execute code in a sandboxed environment
• Interact with external APIs

You can enable or disable skills per agent from the agent’s settings or via the /skill command.

Memory

Memory gives your agent persistent context that carries across conversations. There are two types:

• Short-term memory – conversation context within a single session
• Long-term memory – facts and preferences stored permanently in the agent’s knowledge base

Agents can remember user preferences, project context, previous decisions, and any other information that helps them be more useful over time. Memory is stored per-agent and is searchable via semantic search.

Schedules

Schedules let your agent perform tasks automatically on a recurring basis using cron expressions. Common use cases:

• Daily standup summaries at 9 AM
• Weekly report generation every Monday
• Periodic monitoring of external services
• Scheduled social media posts

Create schedules from the agent’s settings or via the /schedule command. Each schedule consists of a cron expression and a prompt that tells the agent what to do.

Cron expression format: minute hour day-of-month month day-of-week

Guardrails

Guardrails are safety rules that constrain what your agent can and cannot do. There are two types:

• System guardrails – platform-wide rules that apply to all agents (cannot be removed)
• Custom guardrails – rules you define for your specific agent

Examples of custom guardrails:

• “Never share pricing information without manager approval”
• “Always respond in formal English”
• “Do not execute any file deletion commands”
• “Escalate billing questions to a human”

Manage guardrails from the agent’s settings or via the /guardrail command.

Secrets

Secrets are encrypted environment variables available to your agent at runtime. Use them for:

• API keys for external services
• Database credentials
• Authentication tokens
• Any sensitive configuration your agent needs

Secrets are encrypted at rest using AES-256-GCM and are only decrypted inside your agent’s isolated container. They are never exposed in logs, responses, or the UI.

Manage secrets from the agent’s settings or via the /secret command.

MCP Servers

MCP (Model Context Protocol) servers extend your agent’s capabilities by providing additional tools and data sources. Lenno supports two types:

• System MCP – always-on, handles memory, skills, and agent self-configuration
• User MCPs – custom integrations like Fastmail, Microsoft 365, or your own MCP servers

User MCPs run inside your agent’s container for full sandboxing. Configure them from the agent’s MCP Servers tab.