Understanding Sessions
Sessions are conversations and interactions with AI agents across the Lovelace Platform. They maintain continuity whether you're working in the CLI, Desktop app, Web interface, or Mobile—enabling seamless context switching and collaborative work.
What Are Sessions?
A session captures the complete conversation history and interaction context between you and AI agents. This includes messages you send, agent responses, tool usage, file operations, and all the context needed to understand what work is happening.
Sessions provide:
- Conversation continuity: Pick up where you left off, regardless of which Lovelace product you're using
- Context preservation: Agents remember previous discussions, decisions, and work done in the session
- Collaboration history: Complete record of human-agent interactions for review and learning
- Cross-product mobility: Start work in CLI, continue in Studio, finish in Desktop
Cross-Product Sessions
Sessions work identically across all Lovelace products, synchronizing in real-time.
Starting Sessions
CLI: Sessions begin when you start an agent interaction:
lovelace chat # New interactive session
lovelace task "..." # New task-focused session
Desktop: Open a workspace and start a conversation with an agent
Web App: Navigate to a project and initiate an agent interaction
Studio: Create or open a session from the sessions browser
Continuing Sessions
Session synchronization happens automatically. When you start work in one product:
- The session is created in the workspace context
- All interactions are recorded with full context
- Changes sync to other products in real-time
- You can switch products mid-conversation
Example workflow:
- Start debugging in CLI (session created)
- Review code changes in Studio (same session, visual context)
- Commit changes from Desktop (session continues)
- Document findings in Web App (session completes)
The agent maintains full context throughout, understanding the work done in each product.
Session Context Switching
When you switch workspaces or projects, sessions follow workspace boundaries:
- Switch workspace in CLI: Previous session pauses, new workspace session begins
- Open different project in Studio: Each project has its own active sessions
- Navigate workspaces in Web App: Sessions are workspace-scoped
This prevents context bleeding between unrelated work.
Session Types
Lovelace supports three primary session types:
Interactive Sessions
Conversational interactions where you directly communicate with agents:
- Ask questions and get answers
- Request code changes and review suggestions
- Explore ideas through discussion
- Debug problems collaboratively
Interactive sessions feel like talking to a knowledgeable colleague who remembers your entire project context.
Agent Sessions
Autonomous agent execution where agents work independently:
- Agents execute multi-step workflows
- Tool usage happens without user confirmation
- Agents make decisions based on workflow rules
- Progress updates stream to the session
Agent sessions run in the background while you continue other work, with full visibility into agent actions.
MCP Sessions
Model Context Protocol interactions where agents use external tools:
- Connect to MCP servers providing specialized tools
- Discover available capabilities dynamically
- Execute tool calls through MCP
- Track tool usage and results
MCP sessions integrate external functionality seamlessly into agent workflows.
Session Contents
Sessions capture rich interaction details beyond simple message exchange.
Messages
Human and agent messages with full context:
- User questions, requests, and feedback
- Agent responses, explanations, and suggestions
- System notifications and status updates
- Inline code blocks and formatting
Tool Usage
Complete record of agent actions:
- Which tools the agent used
- Input parameters for each tool
- Results and outputs
- Execution time and status
File Operations
Detailed file change history:
- Files read to understand context
- Files created or modified
- Exact changes with line-by-line diffs
- Programming language and syntax
Command Execution
Terminal commands run during the session:
- Commands executed
- Working directory
- Standard output and errors
- Exit codes
Agent Actions
High-level reasoning and planning:
- What the agent is thinking
- Plans before execution
- Verification steps
- Results and conclusions
All of this creates a comprehensive audit trail of work done during the session.
Session Management
Browsing Sessions
Find previous work across all your sessions:
- Filter by workspace: See sessions for specific projects
- Filter by date: Find recent or historical sessions
- Search content: Full-text search across all messages and actions
- Filter by type: Interactive vs agent vs MCP sessions
Session History
Sessions persist indefinitely unless explicitly deleted:
- Full interaction history preserved
- Context searchable and reviewable
- Agents can reference previous sessions
- Export sessions for external analysis
Resuming Sessions
Continue previous work seamlessly:
- Automatic context restoration: Agent remembers entire conversation
- File state preservation: Understands what files were being worked on
- Task continuation: Picks up incomplete work
- Cross-product resume: Open in any Lovelace product
Session Sharing
Collaborate through shared sessions:
- Team sessions: Multiple users in one conversation
- Session links: Share specific sessions with teammates
- Read-only access: Review past sessions without editing
- Session export: Extract for documentation or training
Session Security and Privacy
Workspace Boundaries
Sessions respect workspace permissions:
- Only accessible to workspace members
- Guests see limited session history
- Cross-workspace access requires explicit grants
Session Encryption
All session data is encrypted:
- Messages encrypted in transit and at rest
- Tool outputs encrypted
- File changes encrypted
- Command outputs encrypted
Audit Logging
Session access is logged:
- Who viewed which sessions
- When sessions were accessed
- What actions were taken
- Export and sharing events
Data Retention
Sessions follow workspace retention policies:
- Archived sessions remain accessible
- Deleted sessions are permanently removed
- Export before deletion recommended
- Compliance with organizational policies
Session Best Practices
Descriptive Titles: Give sessions meaningful titles to aid future discovery.
Session Boundaries: Start new sessions for distinct tasks rather than continuing unrelated work in one session.
Regular Review: Periodically review session history to learn patterns and improve agent interactions.
Cleanup Old Sessions: Archive or delete sessions no longer needed to reduce clutter.
Context Building: Reference previous sessions when relevant—agents can learn from past interactions.
Related Concepts
- Agents - How agents participate in sessions
- Memory - How agents remember across sessions
- Workspaces - Organizational boundaries for sessions
- Tools - Actions agents take during sessions