Enterprise Deployment

​

Overview

Devin is designed for seamless integration into enterprise environments, with deployment options that balance speed, security, and compliance. Devin can be initiated through the web interface, Slack, or API, ensuring flexibility in how teams engage with the system. Upon activation, Devin operates within a dedicated workspace that includes:

• A shell for executing commands.
• A browser for web-based interactions.
• A code editor for reading and writing code.

Devin’s workspace operates under the control of its brain, which always resides within Cognition’s Cloud.

---

​

Devin’s Architecture

Devin’s architecture consists of two key components:

• The Brain: A stateless, cloud-based service that powers Devin’s intelligence, always residing in Cognition’s Cloud (similar to GitHub Copilot’s architecture).
• The Devbox: A secure virtual environment where Devin executes code, connects to resources, and interacts with your systems.

The deployment model you choose determines where the Devbox runs and how it connects to your infrastructure.

​

Enterprise SaaS Architecture

In the Enterprise SaaS model, both Devin’s brain and Devbox run in Cognition’s secure, multi-tenant cloud. All data stays encrypted in transit and at rest. Each Devin session runs on its own isolated machine, keeping customer data segregated by default.

​

Customer Dedicated SaaS Architecture

In the Customer Dedicated SaaS model, Cognition hosts Devin in an auto-scaling, customer-isolated environment within a single-tenant VPC. Your VPC connects via AWS or Azure Private Link (or IPSec tunnel), allowing Devin to securely access your privately networked resources. Customer data stays encrypted in transit and at rest, and is processed in an isolated tenant.

​

Customer Hosted VPC Architecture

In the Customer Hosted VPC model, Devin’s Devbox runs entirely within your Virtual Private Cloud. Data is transmitted to Cognition’s tenant for processing by Devin’s Brain, but no customer data is saved at rest outside of your VPC. All customer data is encrypted in motion and at rest within your environment.

---

​

Deployment Options

Devin supports three primary deployment models to meet varying enterprise requirements:

​

Choosing a Deployment Model

Enterprise SaaS Deployment is recommended for most organizations looking for a quick setup with minimal operational overhead. Deployment can be completed within minutes. This model works well when your source code management (GitHub.com, GitLab.com, Azure DevOps Cloud) and artifact stores are publicly accessible or can support IP whitelisting. Customer Dedicated SaaS is ideal for strategic enterprises whose resources are on private networks and cannot support IP whitelisting. In this model, Cognition hosts Devin in an auto-scaling, customer-isolated environment within a single-tenant VPC. Your VPC connects to Cognition’s infrastructure via a secure AWS or Azure Private Link (or IPSec tunnel), allowing Devin to access your privately networked resources while maintaining tenant isolation. This deployment model supports MFA VPN access to your internal resources. Additionally, Customer Dedicated SaaS is typically 20-25% more cost-effective than Customer Hosted VPC deployments. Customer Hosted VPC Deployment provides the highest level of control for enterprises with strict security, compliance, or data isolation requirements. While deployment takes longer, it ensures full network and infrastructure control with zero data retention outside your environment. All customer data is stored in encrypted S3 buckets within your cloud infrastructure. Once a deployment model is chosen, the next critical step is integrating source code repositories.

---

​

Deployment Specifications

​

Customer Dedicated SaaS Requirements

For Customer Dedicated SaaS deployments, Cognition manages the infrastructure on your behalf. Requirements include:

• Network Connectivity:
  • AWS Private Link or Azure Private Link (preferred)
  • IPSec tunnel (alternative option)
  • Ability to establish secure tunnel between your VPC and Cognition’s single-tenant VPC
• Access Configuration:
  • DNS resolution for your internal resources
  • Network routing configured to allow Devin’s Devbox to reach your SCM, artifact stores, and other development tools

​

Customer Hosted VPC Requirements

Deploying Devin in a Customer Hosted VPC requires a dedicated server that meets the following requirements:

• Hardware Requirements:
  • Must support virtualization (Devin does not run in containers).
  • Sufficient compute and memory resources based on usage.
  • AWS: i3.metal instances recommended
  • Azure: Standard_L80as_v3 instances recommended
• Software Requirements:
  • Cognition Agent Service installed on the provisioned server.
  • Secure network connectivity to Cognition’s Cloud.
• Cloud Provider Support:
  • AWS and Azure are fully supported
  • GCP is not currently supported (contact sales for updates)

Best Practices:

• Deploy Devin in a dedicated VPC separate from your main production environment
• Ensure network access to required endpoints: api.devin.ai, *.devin.ai
• Configure DNS to use AWS or Azure DNS resolvers (avoid custom DNS resolvers for easier Private Link setup)
• Use Terraform for infrastructure provisioning (setup scripts are not idempotent)

​

Cross-Tenant Communication

Devin’s architecture ensures secure communication between your environment and Cognition’s Cloud.

---

​

Deployment Guides

Use the following platform-specific guides to deploy Devin in your VPC environment:

---

​

FAQs & Additional Information

---

​

Next Steps

• For Enterprise SaaS Deployment: Start using Devin immediately by logging in to the web app.
• For Customer Dedicated SaaS: Contact our Enterprise Sales Team to discuss your networking requirements and begin the setup process.
• For Customer Hosted VPC Deployment: Follow the AWS or Azure guides to configure your environment.
• Need Assistance? Contact our Enterprise Sales Team.

---