# SurrealDB’s high-speed AI agent context layer > Source: > Published: 2026-07-06 15:50:10+00:00 # SurrealDB’s high-speed AI agent context layer [SurrealDB](https://surrealdb.com/) Is a four year-old UK technology startup developing multi-model database technology combining relational, full-text, document, time-series, graph and vector databases in a single distributed, scale-out database. It has its own in-house storage engine providing real-time access to an organization’s multi-dimensional memory for decision-making and provides a context layer for AU agents The company was founded in 2022 by two brothers, CEO Tobie Morgan Hitchcock and COO Jaime Morgan Hitchcock. Tobie achieved a Masters in Software Engineering from the University of Oxford, with a thesis focused on the use of temporal versioning within distributed graph databases. He is a CEO skilled in database architecture and coding. The company has raised $44 million in funding through a $6 million 2022 seed round and a 2-part A-round with $20 million in 2024 and $18 million in February this year. We asked him what was the trigger that prompted him to produce a multi-model database? He said: "It's the fact that you're using time series databases, document databases, graph databases, key value databases and related databases in every single app. Obviously nowadays you're using it even more, especially graph. But back then it was just like, how can we do something like Firebase, but actually make it a database that can scale to hundreds of terabytes or petabytes of data in a transactional way?" "[And] to work out how to do that in a graph setting, because graph is, I think, the best way of modelling data. How do you do versioning of it? So that was my dissertation at Oxford. It was in temporal versioning, key value store for graph." What Surreal has done is to build its own storage engine: "We've designed our own distributed store. It's not based on Raft. It's based on something called [TAPIR](https://syslab.cs.washington.edu/papers/tapir-tr-v2.pdf), Transactional Application Processing with Inconsistent Replication. We're the first distributed TAPIR store in the world to implement on top of this." He said: "We've built TAPIR into the SurrealDB nodes itself. So SurrealDB is … stateful … but it sits on top of object storage, which is stateless. So you can scale the SurrealDB nodes out to 15 nodes, that's kind of limit for the moment, but those 15 nodes will be running on top of a multi-petabyte scale object storage and we'll be pulling down the data as we need." The in-house storage engine means that the same database can serve relational, full-text, document, time-series, graph and vector database queries without having partitioned copies for each search data type. Some other suppliers are combining database models, such as [Redis](https://www.blocksandfiles.com/ai-ml/2026/05/18/redis-agentic-ai-flowers-with-iris/5241795), [Snowflake](https://www.blocksandfiles.com/ai-ml/2025/11/05/snowflakes-agentic-ai-news-assault/1588549), [Databricks](https://www.blocksandfiles.com/data-management/2026/06/18/databricks-expands-lakehouse-to-unify-olap-and-oltp/5258373) and [SingleStore](https://www.blocksandfiles.com/ai-ml/2025/10/01/singlestore-wants-to-be-chatgpt-for-data-and-ai-apps/1615045). Morgan Hitchcock said: "Yes, but they do that by having two separate databases, right? So Snowflake has OLAP and it's bought, it's launched Postgres layer. Databricks has bought [Neon](https://www.blocksandfiles.com/container-storage/2025/06/16/databricks-rolls-out-lakebase-postgres-and-agent-bricks-for-ai-era-apps/1592030) and they're still completely separate. They're separate databases." And a problem with "Neon is it's single writer. They don't really talk about it as much. Single writer accepts one node writes and you can have a hundred read replicas, but that doesn't scale for an organisation. So it's great to get going, but it doesn't scale. Our design will enable us to have 15 write nodes or 30 write nodes, and then you'll be able to scale out and redirect to hundreds or thousands of those as well." This is serious software and the company has raised serious money, particularly so as it is not based in Silicon Valley and so outside that area's looking glass. SurrealDB is not included in Forrester's recent multi-modal database [report](https://www.blocksandfiles.com/data-management/2026/07/01/the-four-tops-forrester-ranks-multi-model-database-suppliers/5264860) for example. The funding is justified because the company already has two products available; SurrealDB SW, source-available code [on GitHub](https://github.com/surrealdb/surrealdb), and [SurrealDB Cloud](https://surrealdb.com/docs/manage/cloud), a cloud managed service, and is earning revenues from them. SurrealDB’s context memory software has similarities to HYCU’s concept of organizations having to develop [corporate memories](https://www.blocksandfiles.com/data-protection/2026/06/15/hycus-springboard-organizations-will-develop-corporate-memories/5255062) and its graph-based [aiR ](https://www.blocksandfiles.com/data-protection/2026/05/14/hycu-adds-agentic-backup-data-intelligence-layer-to-find-and-fill-risk-gaps/5240353)AI Resiliency product. Unbeknownst to HYCU, SurrealDB has been developing this technology in-house, with a core team of just 3 to 4 people, and already notched up customers such as Apple, BYD, Samsung, Verizon, Tencent, and PolyAI. Verizon used SurrealDB to build a generative AI assistant that serves 10,000 field technicians with instant access to documentation, outage updates, and troubleshooting workflows, improving field operations at scale. It merged relational, document, and graph data into a single platform to simplify retrieval and delivered live outage and service disruption data so technicians can respond faster. Average response times dropped by 40 percent, restoring service faster. Automated knowledge-sharing cut training expenses by more than half while boosting competency. China’s Tencent’s monitoring platform involved nine backend tools and users had to choose the right storage or compute engine before answering any question, slowing investigations. These tools were consolidated into SurrealDB to build a graph-first monitoring platform. A cluster of 9 storage nodes and 6 compute nodes manages 8 million nodes and 50 million edges at 10,000+ QPS. SurrealDB provides fast access to petabytes of data. Samsung Ads uses SurrealDB to build real-time knowledge graphs for campaign execution, unifying content, user, and device data in one layer and replacing three legacy data stores. Query times were reduced from hours to seconds, allowing for real-time campaign adjustments. SurrealDB eliminated the need for multiple analytics platforms, cutting operational costs by 30 percent. More precise audience targeting increased ROI by 25 percent, saving millions in annual ad spend. These are solid customers. How does SurrealDB’s multi-model database work? We’ll explain this in a follow-up article. To whet your appetite watch a YouTube [video](https://www.youtube.com/watch?v=USg8ZQC5mQc&time_continue=601&source_ve_path=MjM4NTE&embeds_referring_euri=https%3A%2F%2Fwww.google.com%2F).