After Two Pivots, Earthly Lands US$5.7M for Its Lunar Guardrails Engine
Earthly, a San Francisco-based provider of a guardrails engine platform for human and AI-generated code, raised US$5.7M in funding. The round was led by Work-Bench, with participation from Innovation Endeavors, Uncorrelated Ventures, Gaingels, and angel investors including Olivier Pomel, Dan Barak, Charlie Marsh, Mengying Li, and Cosmin Nicolaescu. The round was announced on November 17, 2026, and it brought the company’s total funding to US$14.3M. Earthly is led by founder and CEO Vlad Ionescu. Ionescu previously founded Qwiet AI, which was acquired by Harness, and worked on large-scale cloud systems in Google Search and Ads and at VMware. Work-Bench is a New York seed fund whose co-founders and general partners Jessica Lin and Jonathan Lehr lead investment decisions.
The product behind the raise is Lunar. It collects and normalizes evidence from code repositories and CI/CD pipelines, which are the automated systems that build, test and ship software. It then runs deterministic policies at three points: inside AI coding tools, on pull requests, and before deployment. Earthly plans to use the funds to accelerate growth, expand engineering and customer support for enterprise deployments, and extend its instrumentation to more customer tools and workflows.
Market Context
Earthly’s route to this round has been bumpy. The company earlier raised a US$6.5M seed-plus round led by Innovation Endeavors, which brought its total to US$8.6M at the time. Its build-tool products did not hold up commercially. Earthly CI shut down in October 2023 after failing to convert adoption into revenue, and Earthly Cloud and Satellites were discontinued on July 16, 2025. Ionescu then interviewed more than 100 businesses. Earthly says the pivot is gaining traction: Lunar’s enterprise footprint now spans more than 100,000 repositories, and one global communications company has deployed it across more than 11,000 repos under a two-year commitment.
Demand for this kind of tool comes from the fast growth of AI-written code. In organizations that have adopted AI coding tools, AI now authors roughly half of all committed code, according to Veracode. Veracode’s 2026 report found that roughly 44% of AI code generation tasks introduced a risky security vulnerability, and the average security pass rate of 56% has barely changed from 55% in its first report. Earthly competes for platform-team budgets with internal developer portals and scorecard tools such as Backstage, Cortex, OpsLevel and Port, as well as with security scanners. Earthly’s pitch is that those tools report on compliance after the fact, while Lunar blocks or flags a change at the moment it is proposed.
Key Signal
“…consistent automated checks will become as fundamental as monitoring the software in production.”
— Mirko Novakovic, CEO of Dash0, in Earthly’s funding announcement

Regional Relevance
For the United States, Earthly is betting that AI coding tools have made human review too slow to be the main quality check, so standards must be enforced automatically. That bet matters for U.S. enterprises that are pushing AI coding adoption while carrying unresolved risk. Earthly’s customer examples point to heavily regulated U.S. sectors. They include a financial services team that turned 65 pages of strict delivery policy into automated guardrails, and a defense technology company that had spent weeks verifying compliance and preparing evidence for Department of Defense deliveries.
Earthly markets more than 200 built-in guardrails mapped to the CMMC, NIST and STIG security frameworks. That positions it in the federal contractor compliance market as well. For the region’s developer-tools ecosystem, the round shows that AI governance can attract fresh capital even to a company whose earlier products shut down.
The Other Side
Can a seed-sized company win a category that larger players may absorb? Security vendors and developer portal companies are moving toward AI code governance, and US$14.3M in lifetime funding is small next to them. Earthly’s defense is distribution. A footprint of 100,000+ repositories and multiyear contracts could make Lunar hard to remove once it is embedded in enterprise workflows. Customer names, however, have not been disclosed.
Do rigid rules slow teams down? Deterministic checks are predictable, but too many blocking rules could recreate the friction that AI tools were meant to remove. Earthly’s answer is gradual rollout. One hiring-technology customer first used Lunar to assess adherence, then moved to alerting developers, and only turned on blocking as teams were ready.
How reliable are the company’s own figures? Pomegra noted that Earthly has given inconsistent public figures for its prebuilt guardrail library, citing both 100+ and 300+ guardrails in different materials. The difference may simply reflect product growth over time. Still, after two discontinued products, enterprise buyers will likely want proof of staying power, such as revenue disclosures or named reference customers.