Overview
Ona turned years of cloud development infrastructure into a place where AI agents could keep working after a developer closed a laptop. It began as Gitpod, a TypeFox project released publicly in 2018 and separated into a company in 2020. OpenAI announced its acquisition on June 11, 2026; Ona confirmed completion on August 10. The team joined Codex. [1] [10] [11]
Its distinctive asset was the execution environment around the model: code, tools, credentials, isolation, and company controls. That asset gained strategic value as coding agents needed longer running sessions. The standalone cloud service had a different outcome. On October 13, Ona announced that Ona Cloud would close on October 31, with new registrations already stopped. [12]
Image 1 / 1
Founding Story
Gitpod came from people who had already built developer tools together. TypeFox traces its roots to Xtext in 2008, its own establishment in 2016, and the start of Theia in 2017. Its company history places Gitpod’s public launch in 2018 and the separation into a company in 2020. This distinguishes product origins from company separation. [1]
General Catalyst’s 2021 investment account names Sven, Johannes, Moritz, and Jan as founders. Wellfound supplies the full names Sven Efftinge, Moritz Eysholdt, and Jan Köhnlein; OpenAI identifies Johannes Landgraf. Its thesis centered on describing a development environment in code, then recreating it without manual machine setup. [2] [14] [10]
In TypeFox’s August 2020 launch announcement, Efftinge said: “Think CI/CD applied to dev environments.” He also said: “Gitpod allows developers to start coding instantly”. The comparison was practical. Continuous integration already prepared software automatically after changes; Gitpod applied that approach to the computer a developer needed before starting work. [3]
Leadership evolved with the business. The company’s 2022 funding announcement identified Landgraf as sole CEO and Efftinge as its first Technical Fellow. The later agent product reused the environment foundation while changing who performed the work. [4]
Timeline
- 2018: TypeFox released the first public Gitpod version. [1]
- 2020: Gitpod became a separate company and open sourced its environment platform. [1] [3]
- 2022: A company announcement disclosed a $25 million round led by Tom Preston-Werner. [4]
- October 2024: Gitpod explained its move away from Kubernetes and the introduction of Gitpod Flex. [5]
- September 2, 2025: Gitpod adopted the Ona name and emphasized software engineering agents. [6]
- June–August 2026: OpenAI announced the acquisition, then Ona confirmed its completion. [10] [11]
- October 31, 2026: Scheduled closure of Ona Cloud, announced on October 13. [12]
What They Built
Gitpod prepared a disposable development computer from a repository. It fetched dependencies, built the project, and opened the selected branch in a ready environment. Repository configuration replaced instructions that each developer otherwise repeated. Its early distribution crossed GitHub, GitLab, and Bitbucket rather than requiring one source hosting platform. [3]
Ona added agents that could perform tasks inside those environments. A user could delegate work, inspect progress, review changes, and move between a conversation and an editor. The product paired execution with enterprise controls, including private deployments, scoped access, and activity records. [6]
The infrastructure was hard won. Gitpod’s engineers described interactive workloads that need sudden CPU bursts, broad system permissions, and reliable preservation of unfinished work. Packing them together created resource conflicts; moving storage could delay startup. Gitpod Flex replaced much of that arrangement with infrastructure designed for development environments. [5]
Market Position
Target Customers
The original product reached individual developers through repositories and open source. Enterprise adoption required a different sale: reducing setup work while keeping source code and access inside company boundaries. Kingland’s published account describes that progression from standardized environments to parallel agent tasks. [7]
Market Size
No reliable revenue total or addressable market estimate was established. OpenAI said Ona had helped two million developers. That cumulative reach does not establish paid accounts, active use, or recurring revenue. [10]
Competition
GitHub Codespaces already supplies cloud development environments and repository based configuration. Ona’s position increasingly depended on customer controlled deployment and the orchestration around agents. The model suppliers then became both partners and potential owners of the surrounding workflow. [13] [10]
That relationship carried a concrete risk. In August 2026, Landgraf said Anthropic intended to end Ona’s model access because of the planned acquisition. Ona removed its managed Anthropic option and moved existing automations to OpenAI models. Customer supplied Anthropic keys were still supported at the announcement, with future availability uncertain. This is Ona’s account; no matching Anthropic statement was established. [9]
Business Model
Published pricing material listed a Core plan from $20 per month, with pooled usage credits rather than a charge for each seat. Enterprise pricing was negotiated and included private infrastructure and stronger administration. The live pricing address now leads to a sales form; its older offer should not be read as current signup availability. [8]
The economic tension was compute consumption. Idle state costs money, while developers expect fast startup and responsive tools. The engineering account explicitly describes the conflict between dense machine usage and predictable performance. Public evidence does not establish gross margin, profitability, or acquisition consideration. [5]
Traction
General Catalyst reported more than 350,000 developers in 2021. By the rebrand, Ona claimed two million developers and sharply growing enterprise business. These are company and investor statements, without independently verified retention or revenue. [2] [6]
Kingland gives a more concrete usage example. Its company hosted case study describes agents handling dependency migrations, documentation, and refactoring inside its AWS setup. The account supports a real workflow and named customer, while remaining promotional evidence. [7]
Post-Mortem
The infrastructure became more valuable when the worker changed
The central mechanism is an inference from the product history and acquisition rationale. Infrastructure built to spare humans from setup also gave agents a controlled place to execute work. The move to agent execution made its isolation and deployment work relevant to OpenAI’s stated enterprise goal.
In OpenAI’s June announcement, Landgraf said: “Agents need more than intelligence”. OpenAI’s stated purpose was persistent execution under enterprise controls. That directly matches the environment problem Gitpod had pursued before agent products became its focus. [10]
Strategic fit did not preserve every customer product
The strongest alternative explanation is simple consolidation: OpenAI could combine a useful product with Codex distribution. The closing statement supports that rationale. It does not reveal whether independent operation was financially sustainable or what investors received. [11]
Customers faced a separate continuity problem. Ona Cloud environments would not move automatically to Codex. Users had to save work and establish a new environment; October payments and unused purchased credits would be refunded. The notice concerns Ona Cloud and does not establish termination of every private enterprise contract. [12]
Supplier choice needed an operational fallback
The Anthropic access change exposed how a nominally flexible agent environment could still depend on one commercial relationship. Ona could change the model behind automations, but customers also needed confidence about behavior, permissions, and results after that switch. Model portability and work continuity are separate requirements. [9]
Key Lessons
- Gitpod’s early repository integration made environment setup repeatable. Ona later applied that foundation to agents, preserving useful engineering across a product change.
- Its Kubernetes experience shows why startup speed and resource density need separate measurements. Efficient packing can produce an unresponsive development computer.
- OpenAI’s acquisition and Ona Cloud’s closure describe different outcomes. Buyers should evaluate the product they rely on alongside the strategic value of its team and technology.
- The model access change and migration deadline make recovery a product requirement. A saved repository alone does not capture every credential, environment setting, or unfinished task.