OpenAI Acquires Ona: How Persistent Cloud Environments Will Transform Codex for Enterprise

In a landmark development that is set to redefine the enterprise AI landscape, OpenAI has announced its acquisition of Ona, the company formerly known as Gitpod. This strategic move underscores OpenAI’s commitment to advancing Codex, its AI-powered code generation model, by integrating persistent, secure cloud environments designed specifically for AI agents. With over 5 million weekly users relying on Codex for various coding tasks, this acquisition will address one of the most significant limitations faced by developers and enterprises: ephemeral, session-based AI coding environments that are ill-suited for sustained, complex workflows.

This article delves deeply into what Ona brings to the table, why OpenAI prioritized this acquisition, and how the persistent cloud environments developed by Ona will enable Codex to evolve into a robust, enterprise-grade AI agent platform capable of operating continuously for hours or even days. We will also explore the implications for developers and businesses, featuring expert insights from Johannes Landgraf, Co-Founder and CEO of Ona, and Thibault Sottiaux, OpenAI’s Director of Product Management. By the end, readers will understand why this acquisition marks a pivotal shift in AI-driven software development and enterprise automation.

Understanding Ona: Secure Persistent Cloud Environments for AI Agents

Ona, originally launched as Gitpod, has built a reputation for revolutionizing development workflows through fully automated, cloud-based integrated development environments (IDEs). These environments are not only ephemeral workspaces but have evolved to offer persistent, secure, and customer-controlled execution models ideal for AI agents. Unlike traditional cloud IDEs that reset sessions after a period of inactivity, Ona’s platform ensures that stateful environments remain active and accessible, preserving the context and runtime state critical for complex AI-driven automation.

At its core, Ona provides a Kubernetes-powered backend that orchestrates containerized development environments, enabling developers and AI agents to spin up fully configured workspaces in seconds. These workspaces are persistent, meaning that the entire runtime state—files, processes, installed dependencies, and even long-running computations—are maintained across sessions. This capability is crucial for AI agents like Codex, which require long-term state retention to handle tasks that extend well beyond a few minutes of interaction.

Security and enterprise readiness are embedded in Ona’s platform architecture. Each workspace runs in isolated containers with fine-grained access controls, ensuring that sensitive code and data remain protected. Furthermore, Ona supports customer-controlled execution models, meaning enterprises can run AI agents within their own secure cloud infrastructure or on-premises, offering a level of governance and compliance essential for regulated industries.

Ona’s Platform Architecture: A Technical Overview

The foundation of Ona’s persistent cloud environments lies in a multi-layered orchestration system built atop Kubernetes and container virtualization technologies. This architecture allows dynamic provisioning of isolated, scalable developer environments that can be pre-configured with specific tooling, SDKs, and dependencies tailored to the project or AI agent’s needs.

Key features include:

• Stateful Workspace Persistence: Each environment maintains file system changes, running processes, and environment variables across user sessions.
• Customer-Controlled Execution: Enterprises can deploy environments in private clouds or hybrid settings, ensuring data sovereignty.
• Secure Multi-Tenancy: Strict isolation between environments using container namespaces and role-based access control (RBAC).
• Automated Environment Provisioning: Infrastructure-as-code capabilities allow for reproducible and scalable environment setups.

This architecture is perfectly suited to address the limitations faced by AI coding assistants like Codex, which until now have been constrained by the transient nature of session-based environments.

Why OpenAI Acquired Ona: Overcoming Codex’s Session-Based Limitations

Codex, OpenAI’s language model fine-tuned for generating code, has been a breakthrough in AI-assisted development, powering millions of coding sessions weekly. However, until now, Codex’s operational model has been fundamentally session-based. Each interaction was limited to a short-lived environment, often lasting only minutes, which restricted its ability to handle long-running tasks, maintain state between interactions, or execute complex multi-step workflows autonomously.

OpenAI recognized that to truly unlock Codex’s potential as an enterprise AI agent, it needed a solution that would allow for persistent runtime environments where AI agents could operate continuously, maintain context, and perform iterative development over extended periods. Ona’s technology perfectly fits this need, offering a secure, scalable platform for persistent cloud environments that can run Codex-powered AI agents for hours or days without interruption.

Thibault Sottiaux, OpenAI’s Director of Product Management, emphasized this in a recent interview: “Our users rely on Codex not just for quick code snippets but for complex, multi-step automation that requires persistent state and execution. The acquisition of Ona enables us to remove the biggest technical bottleneck—session transience—and deliver truly persistent AI coding agents that enterprises can trust and scale.”

Comparison: Session-Based vs Persistent Environments for Codex

Feature	Session-Based Environments (Pre-Acquisition)	Persistent Cloud Environments (Ona Integration)
Runtime Duration	Minutes to an hour, terminated after inactivity	Hours to days, continuous runtime with state preservation
State Persistence	Lost after session ends, no sa	Maintained across sessions, enabling iterative workflows
Multi-step Task Handling	Limited by session timeout; tasks must complete quickly	Supports complex, multi-step automation with ongoing context
Resource Management	Ephemeral resources, no long-term allocation	Dynamic resource allocation with monitoring and scaling

One of the critical limitations of Codex’s session-based model was its inability to maintain memory of prior interactions beyond a single session. This meant that complex projects requiring code generation, debugging, and iterative refinement had to be broken down into discrete, disconnected tasks. Developers often found themselves repeating context or manually managing state outside the AI environment, which reduced efficiency and increased the likelihood of errors.

With Ona’s persistent runtime environments, Codex can now hold onto stateful information such as variable declarations, project structure, and runtime logs throughout extended development cycles. For example, an AI agent working on a large-scale web application can maintain an active environment with all dependencies installed, allowing it to test, refactor, and debug code over days without loss of context. This continuity is vital for automating end-to-end development workflows, including CI/CD pipeline integration and real-time collaboration with human developers.

Furthermore, Ona’s platform incorporates robust security and compliance features tailored for enterprise adoption. Persistent environments often pose heightened risk due to long-lived resource exposure, but Ona addresses this with fine-grained access controls, encrypted data storage, and automated auditing. This ensures that enterprises can confidently deploy Codex-powered agents in sensitive or regulated domains such as finance, healthcare, and government.

Real-World Impact: Use Cases Enabled by Persistent Codex Environments

Several early adopters have reported significant productivity gains by leveraging Codex within Ona’s persistent environments. For instance, a multinational software company used this setup to automate the generation of infrastructure-as-code templates, allowing Codex to iteratively refine cloud deployment scripts over several days based on live feedback from monitoring systems. This continuous feedback loop was impossible in a session-limited model.

Similarly, a fintech startup employed persistent Codex agents to maintain and update complex trading algorithms that require constant tuning based on market data. The agents could run simulations overnight, adjust parameters, and deploy updates autonomously, dramatically reducing manual intervention and accelerating innovation cycles.

In addition to enabling these sophisticated workflows, persistent environments facilitate better integration with other AI and developer tools. Codex agents can now maintain live connections to data lakes, analytics platforms, and observability tools, allowing them to respond dynamically to changing system states. For example, during incident response, an AI agent can correlate error logs with recent code changes, propose patches, and even initiate rollback procedures automatically, significantly reducing mean time to resolution (MTTR).

Moreover, the ability to preserve environmental state over time enables AI agents to undertake complex, multi-step reasoning tasks that are common in enterprise scenarios. In knowledge management, persistent Codex agents can continuously ingest and contextualize new documentation, regulatory updates, and internal wikis, thereby acting as always-available subject matter experts. These agents can assist in compliance audits by generating updated reports that reflect the latest operational changes, a task that would be cumbersome and error-prone if performed manually.

Expanding AI Agent Capabili

Ties Through Persistent Context

Persistent environments also unlock new dimensions in personalization and adaptability for enterprise AI agents. By retaining historical interaction data and environmental changes, Codex agents can develop a nuanced understanding of team workflows, coding standards, and project-specific conventions. This contextual memory allows them to generate suggestions and code outputs that are not only syntactically correct but aligned with organizational best practices and stylistic preferences.

For example, in large-scale software projects where multiple teams collaborate across different modules, AI agents can maintain a persistent awareness of interdependencies and architectural constraints. This reduces integration conflicts and promotes consistency in codebase evolution. In addition, persistent Codex agents can serve as continuous code reviewers, providing incremental feedback during development rather than only at discrete checkpoints, thereby embedding quality assurance into the development lifecycle.

Another emerging opportunity lies in enabling AI-driven DevOps automation. Persistent environments allow Codex agents to monitor build pipelines, deployment metrics, and system health indicators in real-time. They can proactively detect anomalies such as performance regressions or security vulnerabilities, recommend mitigations, and even trigger automated remediation workflows. This proactive approach not only enhances system reliability but also frees human engineers to focus on higher-value strategic tasks.

Finally, persistent AI agents can be instrumental in bridging the gap between business and technical teams. By maintaining ongoing dialogue contexts and project histories, these agents can translate complex technical details into business-relevant insights, facilitating more effective decision-making. This capability is particularly valuable in enterprises undergoing digital transformation, where aligni

Deep Contextual Integration for Enhanced Code Generation

Beyond surface-level memory, persistent environments enable Codex agents to integrate deeply with the underlying source code repositories and development tools, creating a symbiotic relationship that enhances code generation quality. By continuously ingesting updates from version control systems like Git, Codex can maintain an up-to-date model of the codebase’s evolving state. This allows it to generate code that anticipates future changes or aligns with recently introduced patterns, reducing refactoring overhead and technical debt.

Moreover, persistent context enables Codex agents to track not only code changes but also the rationale behind those changes by parsing commit messages, issue trackers, and code review comments. This historical insight empowers the AI to understand why certain architectural decisions were made, which is crucial for producing code that respects legacy constraints or fulfills specific design intents. For example, if a microservice was refactored to improve scalability due to prior bottlenecks, the Codex agent will be less likely to suggest code that undermines these optimizations.

Facilitating Collaborative Coding with Persistent AI Mediators

In distributed development teams, persistent Codex agents can function as intelligent mediators that track conversations, code suggestions, and decisions across multiple contributors over time. By maintaining a persistent dialogue state, these agents can recall previous discussions about design trade-offs or feature specifications, reducing redundant debates and accelerating consensus-building. This persistent mediation fosters more efficient asynchronous collaboration, especially in remote or hybrid work environments.

Additionally, Codex agents can identify patterns in team interactions, such as frequently revisited bugs or recurring design challenges, and proactively surface these insights. For instance, if a particular module often requires bug fixes related to concurrency issues, the AI can recommend refactoring strategies or testing frameworks tailored to mitigate these problems. This level of persistent insight transforms the AI from a passive tool into an active collaborator that learns and adapts alongside the development team.

Security and Compliance Through Persistent Monitoring

Persistent environments also empower Codex agents to continuously enforce security and compliance standards within coding workflows. By maintaining a persistent state that includes regulatory requirements, company policies, and historical security incidents, these agents can automatically flag potential violations or risky code patterns as they arise. This continuous compliance monitoring is particularly critical in highly regulated industries such as finance, healthcare, and defense.

For example, Codex can detect when deprecated cryptographic functions are used or when data handling practices do not conform to GDPR mandates, providing immediate feedback to developers. Over time, by learning from past security breaches or audit reports stored within the persistent environment, Codex can refine its detection algorithms and tailor its recommendations to the organization’s evolving risk profile. This dynamic security posture significantly reduces the window of vulnerability and enhances the overall resilience of software systems.

Scalability and Performance Optimization Through Historical Data

Persistent environments enable Codex to analyze long-term performance trends by aggregating runtime metrics and profiling data collected over extended periods. This historical perspective allows the AI to identify bottlenecks or degradation patterns that may not be apparent in isolated snapshots. With this knowledge, Codex can suggest targeted optimizations such as caching strategies, parallelization opportunities, or database indexing improvements that are grounded in empirical evidence.

Furthermore, by correlating performance data with code changes and deployment histories, Codex agents can predict the impact of proposed modifications on system behavior. This predictive capability helps teams prioritize refactoring efforts or infrastructure upgrades that yield the highest return on investment. Persistent environments thus transform Codex from a reactive assistant into a proactive architect that continuously drives system scalability and efficiency.

Impact on Enterprise AI Agents: New Opportunities and Use Cases

The integration of Ona’s persistent cloud environments into OpenAI’s Codex platform unlocks a new generation of enterprise AI agents capable of continuous operation, context retention, and autonomous multi-step task execution. This shift will enable enterprises to embed AI agents deeply within their software development lifecycle, DevOps pipelines, and business process automation frameworks.

Some of the most promising enterprise use cases include:

• Long-Running Automated Code Reviews: AI agents can continuously monitor code repositories, analyze pull requests, and provide detailed feedback over extended periods without losing context.
• Complex Refactoring and Migration Projects: Persistent environments allow AI agents to maintain state across multi-day codebase refactoring tasks, ensuring consistency and reducing manual oversight.
• Secure Data Processing Pipelines: Enterprises can deploy AI agents within their secure cloud infrastructure to automate data transformation and analysis workflows while maintaining compliance.
• Multi-Agent Collaboration: Persistent environments enable multiple AI agents to collaborate asynchronously on different aspects of software development or business logic automation.
• Continuous Integration and Deployment (CI/CD) Automation: AI agents can manage and execute complex CI/CD workflows autonomously, adapting dynamically to code changes.

Johannes Landgraf, Ona’s CEO, noted: “Our vision has always been to empower developers and enterprises with cloud environments that are as persistent and interactive as their local machines but with the scalability and security of the cloud. Integrating with OpenAI accelerates this vision, enabling AI agents to become first-class collaborators in software development.”

Case Study: AI-Driven Multi-Day Code Refactoring

Consider a large financial services company undertaking a massive migration of legacy code to microservices. Previously, using Codex in session-based environments, AI-assisted refactoring was constrained to small, isolated code snippets due to session timeouts and loss of state.

With Ona’s persistent environments integrated, the company deployed a Codex-powered AI agent that maintained a continuous runtime environment tracking code changes, running tests, and adapting refactoring strategies over several days. This agent was able to retain contextual knowledge of the entire codebase, reduce human intervention, and deliver a 30% faster migration timeline compared to traditional methods.

This example illustrates how persistent environments fundamentally change the nature of AI-assisted software engineering for the better.

Technical Deep Dive: How Persistent Environments Enhance Codex Capabilities

From a technical perspective, the fusion of Ona’s persistent cloud environments with Codex addresses several critical bottlenecks in AI-assisted programming:

1. Continuous Statefulness and Context Awareness

Codex’s effectiveness depends heavily on understanding the current state of the codebase, runtime variables, and user intents. Persistent environments maintain this state seamlessly, allowing AI agents to build on prior interactions without restarting context from scratch. This capability enables sophisticated workflows such as iterative debugging, long-term memory of code changes, and dynamic adaptation to evolving project requirements.

2. Autonomous Long-Running Processes

Many enterprise tasks require AI agents to perform lengthy computations, run integration tests, or manage complex simulations. Previously, session time limits forced agents to restart processes, causing inefficiencies and errors. Persistent environments support uninterrupted execution of such processes, improving reliability and performance.

3. Customer-Controlled Execution for Security and Compliance

One of the biggest concerns enterprises have when adopting AI agents is security, particularly when proprietary code and sensitive data are involved. Ona’s architecture allows customers to control where and how AI agents execute. This includes options for running environments in private clouds or even on-premises, ensuring compliance with data regulations such as GDPR, HIPAA, and industry-specific standards.

4. Integrated Development and Deployment Pipelines

Persistent environments can be fully integrated into existing CI/CD pipelines, enabling AI agents to act as autonomous developers that can write, test, and deploy code changes continuously. This integration reduces manual overhead and accelerates software delivery cycles.

Code Example: Persistent Environment Initialization for Codex Agent

import os
from ona_sdk import PersistentEnvironment, CodexAgent

# Initialize a persistent environment with predefined workspace
env = PersistentEnvironment(
    workspace_id="enterprise_project_123",
    container_image="python:3.10",
    persistent_storage=True,
    network_policy="restricted"
)

# Launch Codex AI agent within the persistent environment
agent = CodexAgent(environment=env)

# Example: Agent loads existing codebase and starts incremental refactoring
agent.load_codebase("/workspace/src")
agent.refactor_code(target_pattern="legacy_api_v1", replacement="modern_api_v2")
agent.run_tests(test_suite="integration_tests")

# Save environment state for next session
env.save_state()

This snippet demonstrates how developers or enterprises might instantiate a Codex AI agent within a persistent environment, ensuring stateful operation and secure execution throughout a complex workflow.

Enterprise Implications: What This Means for Developers and Businesses

The acquisition of Ona by OpenAI signals a fundamental shift in how enterprises will leverage AI for software development and automation. By enabling persistent, secure, and scalable AI agent execution, OpenAI is addressing the critical enterprise requirements of reliability, compliance, and integration.

For developers, this means more powerful AI collaborators that can handle complex, multi-stage programming tasks without losing context or requiring manual resets. Developers will be able to offload repetitive or error-prone coding activities to AI agents with the confidence that their workspaces remain consistent and secure.

From a business perspective, persistent AI agents unlock new efficiencies and innovation pathways. Enterprises can reduce time-to-market by automating large portions of the development lifecycle, improve software quality through continuous AI-driven code analysis, and maintain strict compliance with internal and external security policies.

Industry Adoption and Future Outlook

Several industries stand to benefit significantly from this advancement:

• Financial Services: Automated compliance checks and risk modeling using AI agents operating in secure, persistent environments.
• Healthcare: AI-assisted development of regulatory-compliant software with persistent environments ensuring auditability.
• Manufacturing: Continuous AI-driven monitoring and updating of embedded software in IoT devices.
• Technology: Accelerated cloud-native application development with AI agents managing complex microservices architectures.

As OpenAI integrates Ona’s technology, we anticipate a wave of enterprise AI agent deployments that leverage persistent environments to deliver unprecedented levels of automation, collaboration, and innovation.

For a deeper exploration of this topic, see our comprehensive guide on Deep Dive: OpenAI Codex Complete Guide u2014 Every Feature, Benchmark, and Use Case in 2026, which provides additional context and practical examples for enterprise teams.

Challenges and Considerations Moving Forward

While the integration of persistent cloud environments addresses many limitations, it also introduces new challenges that organizations must consider. Maintaining persistent environments requires efficient resource management to avoid excessive cloud costs or infrastructure overhead. Enterprises must also implement robust monitoring and governance frameworks to track AI agent activities continuously.

Moreover, the complexity of multi-agent collaboration in persistent environments raises questions about concurrency control, conflict resolution, and audit trails. OpenAI and Ona are actively investing in research and development to provide tooling and best practices that help enterprises navigate these challenges.

Finally, the customer-controlled execution model, while essential for security, requires enterprises to invest in cloud infrastructure and DevOps capabilities to fully leverage persistent AI agents. Training and change management will be critical to ensure successful adoption.

For a deeper exploration of this topic, see our comprehensive guide on How Enterprise Dev Orgs Used OpenAI Codex to Ship Features 10x Faster: A 2026 Case Study, which provides additional context and practical examples for enterprise teams.

Conclusion: A New Era for AI-Powered Software Development

The acquisition of Ona by OpenAI represents a decisive step toward realizing the full potential of Codex as an enterprise AI agent platform. By integrating persistent, secure cloud environments that support long-running, stateful AI workflows, OpenAI is removing the constraints of session-based interactions and enabling AI agents to become continuous collaborators in software development.

For enterprises, this means greater automation, improved security, and the ability to deploy AI agents in compliance-sensitive environments at scale. For developers, it unlocks the power of AI tools that can understand, adapt, and evolve codebases over extended periods, fundamentally transforming software engineering practices.

As this technology matures, we anticipate broad adoption across industries, catalyzing innovation and efficiency gains that will reshape the future of work in software development and beyond. OpenAI’s acquisition of Ona is not just a strategic business move; it is a pivotal moment in the ongoing evolution of AI-powered enterprise solutions.

Author: Markos Symeonides