Eon launches a simulated enterprise for testing agents across business software

Eon launched Era on October 5th, a free environment that generates a simulated company across business software and cloud systems so developers can test AI agents without exposing real company data, co-founder and CEO Ofir Ehrlich said. Era generates linked company histories across Salesforce, Slack, Jira, Zendesk, Gong, SharePoint and Google Drive, with employees, customers and deals represented consistently across systems and deliberate data defects such as duplicate records and inconsistent formatting. Eon's documentation states the synthetic environment is no substitute for a staging system with real customer data, so production integration and safety checks require separate testing.

Eon launches a simulated enterprise for testing agents across business software Era generates linked company data across systems including Salesforce, Slack and Jira, giving developers a repeatable test environment without using production records. By Ryan Merket https://runtimewire.com/author/ryan-merket · Published Primary source: X https://x.com/OfirEhrlich/status/2107124107464905058 Why it matters Enterprise agents need testing that reflects permissions, records and workflows across multiple systems. Era offers developers a repeatable synthetic environment with known ground truth, while its emulated APIs still leave production testing as a separate requirement. Eon launched Era on October 5th, a free environment that generates a simulated company across business software and cloud systems so developers can test AI agents without exposing real company data. Co-founder and CEO Ofir Ehrlich https://x.com/OfirEhrlich said the environment lets agents operate through live MCP servers and vendor-compatible APIs, with the same employees, customers and deals represented across systems. https://x.com/OfirEhrlich/status/2107124107464905058 https://x.com/OfirEhrlich/status/2107124107464905058 Era addresses a problem in enterprise-agent development: testing against real company records raises privacy and security risks, while isolated mock data can miss the messy links agents must handle in production. Eon's Era console https://console.era.eon.io/ generates company histories across tools such as Salesforce, Slack, Jira, Zendesk, Gong, SharePoint and Google Drive. Its website describes employees who appear consistently across systems, deal and ticket histories, and deliberately introduced data defects such as duplicate records and inconsistent formatting. That shared state is central to the product. A test can ask an agent to investigate a customer or resolve a support issue using more than one system, then compare the agent's answer or actions with the exact underlying records. Eon says developers can use the same environment to test, benchmark and improve agents, including using observed failures for targeted post-training and rerunning the environment to measure changes. Era is available through a browser console and a command-line interface, and the product site lists integrations for agent-development tools including Claude Code, Devin, Cursor and Codex. Individual system images can also be run in CI, according to the Era site https://console.era.eon.io/ . Eon says the service is free for builders; it has not announced Era-specific revenue or customer figures. Era emulates systems; it does not provide production instances or complete replicas of every vendor feature. Eon's product documentation says the synthetic environment is no substitute for a staging system containing real customer data, and that the emulators implement the interfaces clients use rather than every part of each vendor API. Developers can use Era for repeatable development and evaluation, but production integration and safety checks require separate testing. Ehrlich's background includes infrastructure testing. He co-founded CloudEndure and led its research and development before Amazon acquired the company in 2019; AWS later incorporated CloudEndure technology into its migration and disaster-recovery products. AWS's account of CloudEndure https://aws.amazon.com/blogs/aws/cloudendure-highly-automated-disaster-recovery-80-price-reduction/ confirms the acquisition, while Vine Ventures https://vineventures.com/blog/big-news-in-the-cloud-eon-launches-out-of-stealth-with-127m-in-funding/ , an Eon investor, says Ehrlich spent about four and a half years at AWS leading those businesses. Eon itself sells cloud backup and data-access software, making Era a move from managing enterprise data to simulating the environments where agents will act on it. In December 2025, Eon said it raised a $300 million Series D led by Elad Gil, bringing its stated total funding to $500 million and its valuation to $4 billion. The round also included returning investors Sequoia Capital, Lightspeed Venture Partners, Greenoaks and BOND, according to Eon's announcement https://www.eon.io/news-and-events/series-d-funding . That funding was for Eon, not a separately disclosed Era round. Era's launch follows two Eon-authored preprints that describe the evaluation problem and the company's approach. In a September 9th benchmark paper https://arxiv.org/abs/2609.09853 , the authors describe generating a coherent enterprise estate with computed answer keys and report a mean realism score of 97 across 23 generated companies, with no records flagged by their synthetic-data detector. A September 24th paper https://arxiv.org/abs/2609.30055 tests agents on questions requiring them to reconcile conflicting evidence across records; its reported results show that the harder questions remained difficult for the agents evaluated. Both are company-authored research, not independent validation of Era's performance. Ehrlich's October 5th post named NVIDIA, Decart https://runtimewire.com/models/fal/decart-lucy-5b-image-to-video , Composio, Openlayer, Deel, Eragon and Plurai as research partners. The partnership list indicates an early focus on developers and AI teams evaluating agents, rather than a standalone enterprise software rollout. The near-term test for Era is whether its generated companies provide enough realistic, cross-system variation to expose failures that mock datasets miss, while remaining consistent enough to grade those failures reliably.