Modern Relay Raises $3 Million for a Shared Context Layer for AI Agents

SAN FRANCISCO, Calif. - Modern Relay has raised $3 million in seed financing to develop infrastructure that gives enterprise AI agents a common source of context. The investment, led by Point Nine, highlights a growing industry focus on the data connectivity layer required to make autonomous software agents functional within complex business environments. Emerge, Amino Collective, and Common Magic also participated in the round, joined by a cohort of technology operators and angel investors who are betting on the company's ability to bridge the gap between large language models and proprietary organizational knowledge.

Founded by Ragnor Comerford and Aaron Goh, the company operates between the technology hubs of Barcelona and San Francisco. This dual-presence strategy allows the firm to tap into both European engineering talent and the concentrated capital and customer base of the Silicon Valley ecosystem. The core of their offering is a platform designed to solve a persistent bottleneck in AI deployment: the tendency for disparate agents to operate in isolation, leading to inconsistent outputs and repetitive data fetching that strains internal systems.

Modern Relay's central product is Omnigraph, an open-source graph database designed to map how information, people, and software relate inside an organization. Unlike traditional flat databases or basic vector stores used for search, a graph database focuses on the relationships between data points. In a corporate setting, this means identifying how a specific contract relates to a certain client, which internal team managed the deal, and what software protocols govern the delivery of the services involved. By mapping these webs of interaction, the company aims to provide a unified reference point for both human employees and digital agents.

The system distinguishes itself by applying versioning concepts familiar from software development, such as those found in Git. This architectural choice is intended to provide a level of transparency that is often missing from contemporary AI applications. By using version control for organizational context, teams can review changes, audit the evolution of data, and understand exactly where a piece of context originated. This lineage is critical for enterprises that are regulated or those that require high levels of accountability for the actions taken by autonomous software.

Modern Relay argues that the current limitations of AI agents are often less a matter of model capability and more a result of fragmented company data. While models from major providers have become increasingly sophisticated in their reasoning, they still struggle when they cannot access high-fidelity, real-time context. A sales agent, for example, may need current product rules, customer history, and specific approval policies that currently live in entirely separate systems. Without a shared context layer, the agent is forced to query these systems individually, often losing the nuance of how those data points interact with one another.

A shared graph can give multiple agents access to consistent relationships and definitions without forcing a company to perform the arduous task of copying all of its data into another vendor's database. This 'in-place' context approach is gaining traction among enterprise architects who are wary of the data gravity created by large cloud providers. By acting as a pointer and relationship mapper rather than a storage repository for the raw data itself, Modern Relay aims to integrate with existing workflows rather than requiring a wholesale migration of company records.

However, this architecture also raises a difficult and non-negotiable requirement: the context layer must respect complex permissions as information changes. In a corporate environment, not all data is equal, and not all users or agents should have total access. Modern Relay acknowledges that enterprises will expect detailed, granular controls over which employees and agents can read or modify a specific record within the graph. This governance layer is often the primary hurdle for startups attempting to sell into the Fortune 500, where data privacy and security are paramount.

Beyond permissions, technical reliability remains a significant barrier to the widespread adoption of graph-based AI infrastructure. Companies will need dependable connectors that can pull data from legacy systems without breaking, as well as a clear recovery process for when an automated update is wrong. If a single agent misinterprets a policy and updates the graph with incorrect information, the error could potentially propagate to every other agent using that graph. These operational details will determine whether Modern Relay's technology becomes foundational infrastructure or simply another data silo that adds complexity without sufficient utility.

The newly secured seed capital will support product development, aggressive hiring across its two primary locations, and the support of early deployments with pilot customers. The funding arrives at a time when 'agentic' workflows are being hailed as the next phase of the generative AI boom. Analysts have noted that while the first wave of AI adoption focused on simple chat interfaces, the next wave is expected to focus on agents that can perform tasks, move between applications, and make decisions on behalf of users—all of which require a deep understanding of organizational context.

Modern Relay is entering a market that is already becoming crowded. Competitors ranging from established enterprise search providers to a new crop of startups focused on agent memory and knowledge graphs are all vowing to solve the same problem. However, Modern Relay's decision to pursue an open-source approach may provide a strategic advantage. It allows technical buyers and developers within these organizations to test the system in their own environments before committing to a commercial contract, potentially lowering the barrier to entry in a skeptical market.

The broader venture capital landscape has shown a distinct appetite for infrastructure plays that support the AI stack, as investors look for 'shovels' in the gold rush of model development. Industry observers suggest that the winners in this space will be companies that can handle the messiness of real-world data—data that is often incomplete, contradictory, and scattered across decades-old software. Modern Relay's focus on relations and versioning suggests an awareness of these structural challenges that simple search-based solutions often overlook.

The ultimate commercial challenge for the company will be turning this technical interest into a repeatable enterprise product with strong, baked-in governance. While open-source adoption can drive initial growth, the path to significant revenue involves proving that the system can scale to tens of millions of nodes without degradation in performance. It also requires convincing IT departments that a new layer of graph infrastructure is worth the overhead of maintenance and oversight.

As Modern Relay begins deploying its seed capital, the industry will be watching to see if its graph-centric model can unify the 'knowledge silos' that have plagued enterprise software for years. If successful, the company could provide the connective tissue that allows a new generation of AI agents to move from mere experimental tools to integrated members of the corporate workforce. The focus now shifts to the startup's ability to execute on its technical roadmap and win the trust of global organizations that are cautious about how they expose their internal data to AI.