Skip to main content

Understanding Deployment

Note: This page explains how the Lovelace Platform itself can be deployed—where your agents run and how the platform infrastructure is hosted. For guidance on deploying applications YOU build with Lovelace, see product-specific deployment guides.

Deployment in Lovelace is about choosing the right execution model for your agents, workspaces, and workflows. The platform supports four deployment models, each optimized for different requirements: local-only, cloud-only, hybrid, and air-gapped.

The Core Insight

Traditional deployment decisions lock you into infrastructure choices early. Lovelace separates the logical definition of agents and workflows from where and how they run. The same agent definition works in local development, cloud execution, or enterprise air-gapped environments.

This means you can develop locally with the CLI, test workflows in Studio, and deploy to Agents Cloud without rewriting anything. Your agents are portable across deployment models.

Deployment Models

Local-Only Deployment

Everything runs on your machine. The CLI, local agent runtime, and file-based storage provide complete Lovelace functionality without any cloud services.

Best for:

  • Individual developers who want full control
  • Sensitive projects requiring data locality
  • Offline development environments
  • Learning and experimentation
  • Open source contributions

Characteristics:

  • Zero cloud dependency: Works without internet connection
  • Local file storage: Data stays on your machine
  • Process-based execution: Agents run as local processes
  • Development speed: No network latency, instant feedback
  • Limited scalability: Bound by local machine resources

What's Local:

  • Agent execution runtime
  • Memory and context storage
  • Workspace definitions
  • Conversation history
  • Repository access

What You Manage:

  • Resource allocation (CPU, memory)
  • Backup and recovery
  • Software updates
  • Security and access control

Example Use Cases:

  • Prototyping new agent workflows before production deployment
  • Working on proprietary code that cannot leave your machine
  • Contributing to open source projects while maintaining privacy
  • Learning Lovelace without cloud infrastructure costs

Cloud-Only Deployment

Everything runs in Lovelace Cloud. Studio, Agents Cloud, and Memory Platform provide fully managed infrastructure with automatic scaling.

Best for:

  • Teams requiring collaboration
  • Production workloads needing reliability
  • Projects requiring significant compute resources
  • Organizations wanting managed infrastructure
  • Global accessibility requirements

Characteristics:

  • Fully managed: Lovelace handles infrastructure, scaling, updates
  • Multi-user: Built-in collaboration and team features
  • Elastic scaling: Resources scale automatically with demand
  • High availability: Redundancy and failover built-in
  • Global access: Available from anywhere with internet

What's Cloud:

  • Agent orchestration and execution
  • Memory Platform storage
  • Workspace registry
  • Collaboration infrastructure
  • Integration management

What Lovelace Manages:

  • Infrastructure provisioning
  • Scaling and performance
  • Backup and disaster recovery
  • Security and compliance
  • Software updates

Example Use Cases:

  • Production agent deployments serving customer workloads
  • Team collaboration on shared projects
  • Resource-intensive tasks requiring significant compute
  • Applications requiring global availability
  • Enterprise deployments needing managed SLAs

Hybrid Deployment

Local development, cloud execution. Develop and test locally with the CLI, deploy production agents to Agents Cloud, design workflows in Studio.

Best for:

  • Development teams following modern practices
  • Organizations balancing control and convenience
  • Projects mixing sensitive and non-sensitive components
  • Gradual migration from local to cloud
  • Cost-optimized architectures

Characteristics:

  • Flexible execution: Choose local vs cloud per agent/workflow
  • Unified experience: Same tools across environments
  • Environment parity: Development matches production
  • Cost control: Cloud resources only when needed
  • Progressive adoption: Start local, move to cloud incrementally

Local Components:

  • Development workspace definitions
  • Testing and experimentation
  • Sensitive credential management
  • Personal context and preferences

Cloud Components:

  • Production agent execution
  • Shared Memory Platform
  • Team collaboration features
  • Long-running workflows

Example Use Cases:

  • Developing agent workflows locally, deploying to cloud for production
  • Keeping sensitive development data local while using cloud compute
  • Team members working locally but sharing cloud infrastructure
  • Cost optimization by using local resources for development
  • Maintaining local fallback when cloud services are unavailable

Air-Gapped Deployment

Self-hosted Lovelace infrastructure. Run Accounts Service, Agents Cloud, and Memory Platform in your own infrastructure with no external dependencies.

Best for:

  • Enterprise security requirements
  • Regulated industries (finance, healthcare, government)
  • Data sovereignty requirements
  • Organizations with strict compliance needs
  • Maximum control requirements

Characteristics:

  • Complete isolation: No external network connections required
  • Full control: Your infrastructure, your rules
  • Custom compliance: Meet specific regulatory requirements
  • Enterprise integration: Connect to internal systems
  • Resource ownership: All data stays within your network

What You Deploy:

  • Accounts Service (authentication and authorization)
  • Agents Cloud (agent orchestration runtime)
  • Memory Platform (persistent context storage)
  • Optional: Developer Portal, Admin Portal

What You Manage:

  • Infrastructure provisioning and scaling
  • Security and network configuration
  • Backup and disaster recovery
  • Software updates and patching
  • Monitoring and alerting

Example Use Cases:

  • Financial institutions with data residency requirements
  • Healthcare organizations managing protected health information
  • Government agencies with security clearance requirements
  • Companies in regulated industries (GDPR, HIPAA, SOC 2)
  • Organizations requiring complete audit trails

Runtime Portability

Agent definitions are runtime-agnostic. The same workflow definition works across all deployment models:

yaml
# This workflow definition works everywhere
workflow:
  name: code-review
  states:
    - name: analyze
      tools: [pathway:git:read, pathway:code:analyze]
    - name: comment
      tools: [pathway:github:comment]

Local Execution: CLI loads definition, runs agent as local process Cloud Execution: Agents Cloud loads definition, runs in container Air-Gapped: Self-hosted Agents Cloud loads definition, runs in your infrastructure

The agent doesn't know or care where it runs. Tool implementations adapt to the environment, but the agent's logic remains identical.

Environment Parity

Lovelace maintains environment parity across deployment models:

Tool Compatibility: Same tools work in local and cloud environments (with environment-appropriate implementations)

Workflow Behavior: State machines execute identically regardless of deployment

Memory Access: Memory APIs work the same, storage backend differs

Authentication: Same auth model, different infrastructure

This parity means:

  • Develop locally with confidence it works in production
  • Test in cloud environments matching production
  • Migrate between deployment models without code changes
  • Debug production issues in local development

Deployment Transitions

Organizations often transition between deployment models:

Local → Hybrid

Start with local-only, add cloud for production workloads:

  1. Develop workflows locally with CLI
  2. Test in local agent runtime
  3. Deploy to Agents Cloud for production
  4. Keep development local, run production in cloud

Local → Cloud-Only

Fully migrate to managed infrastructure:

  1. Export local workspace definitions
  2. Import to Lovelace Cloud
  3. Configure team access and permissions
  4. Migrate memory to Memory Platform
  5. Decommission local infrastructure

Cloud → Air-Gapped

Move to self-hosted for compliance:

  1. Provision infrastructure (Kubernetes cluster, databases)
  2. Deploy Lovelace services (Accounts, Agents Cloud, Memory Platform)
  3. Migrate workspace and agent definitions
  4. Configure internal authentication (SSO, LDAP)
  5. Establish backup and monitoring

Hybrid → Air-Gapped

Maintain local development, self-host production:

  1. Keep CLI for local development
  2. Deploy Agents Cloud in your infrastructure
  3. Configure network boundaries and access control
  4. Migrate cloud agents to self-hosted
  5. Maintain hybrid development workflow

Deployment Best Practices

Start Local, Grow Gradually: Begin with local-only to learn and prototype. Add cloud when collaboration or scaling becomes necessary.

Match Model to Requirements: Choose deployment based on actual needs, not assumptions. Don't pay for cloud if local suffices; don't struggle with local if cloud solves problems.

Test Environment Parity: Ensure development environments match production deployment model. Local development should mirror production architecture.

Plan for Migration: Design with future deployment changes in mind. Workspace definitions should be exportable; agent configurations should be portable.

Monitor Actual Usage: Track resource consumption, costs, and performance across deployment models to inform optimization decisions.

Related Concepts