Building agentic graph RAG and a graph-based memory agent on SurrealDB

Cobrainer's skills-intelligence platform runs a Rust-native agentic graph RAG and a graph-based memory agent on SurrealDB, getting graph, vector, and full-text from one engine instead of bolting on pgvector or OpenSearch.

IndustryHR technology

FocusAgentic graph RAG & agent memory

LocationMunich, Germany

COMPANY OVERVIEW

Who they are

Cobrainer, based in Munich, runs a skills-intelligence platform. As a startup that needed to move fast and iterate at pace, they wanted to improve agent accuracy and token efficiency without relying on a fragmented data stack and having to manage a variety of tools. Cobrainer required a flexible database that can handle many use cases with sufficient ergonomics, not only for engineers but also for the AI agents.

01 |KEY CHALLENGES

What they needed to solve

As Cobrainer looked to add agentic graph RAG and an AI agent with graph-based memory, it wanted these capabilities without fragmenting its stack and adding complexity, getting graph, vector, full-text, and flexible schema from one engine.

Agent accuracy

Flat vector retrieval surfaced only loosely related context. The team wanted graph structure so the agent could follow real relationships and ground its answers more reliably.

Token efficiency

Pulling broad vector matches into every prompt was wasteful, inflating the tokens spent on each agent call.

Agent and graph RAG on one store

An AI agent with graph-based memory living in the database, plus an agentic graph RAG that uses the relations between nodes the agent builds automatically.

Pace of delivery

New capabilities meant standing up new storage patterns quickly, without long migrations slowing the team's pace of innovation.

02 |SOLUTIONS

How SurrealDB helped

SurrealDB gave Cobrainer graph, vector, and full-text in one engine, backing the agentic graph RAG and the agent's graph-based memory directly.

One engine for graph, vector, and full-text

A single store, queried through SurrealQL, covers documents, graphs, vectors, and full-text, so the agentic case needed no pgvector or OpenSearch alongside Postgres.

Rust-native agentic graph RAG

Built natively on SurrealDB's Rust SDK, combining graph traversal and vector similarity in a single query, replacing the earlier S3-and-OpenSearch RAG.

Graph-based agent memory

SurrealDB stores the agent's memory and session checkpoints, with a multi-layered, graph-based memory model that was quick to implement.

Managed cloud inside the EU

SurrealDB Cloud lets a lean team run the engine as a managed service with EU-region residency, keeping HR data inside the required boundary.

As a startup, we didn't want to fragment our stack every time we add a capability. For our agentic graph RAG, the alternative was weighing Postgres with pgvector and search extensions against bolting on OpenSearch, instead we got graph, vector, and full-text from one engine. What surprised us was how quickly we stood up a multi-layered, graph-based memory model for our AI agent, with SurrealDB handling its memory and session checkpoints. And SurrealDB Cloud kept all of it inside our EU data boundary.

Daniel Alker

Senior Cloud Platform Engineer, Cobrainer

03 |RESULTS

The impact

Cobrainer replaced its simple S3-and-OpenSearch RAG with an agentic graph RAG and built an AI agent backed by SurrealDB, adding both without fragmenting its stack. A multi-layered, graph-based memory model was fast to implement, and grounding the agent in graph relationships rather than loose vector matches improved accuracy while cutting LLM token costs per call. Cobrainer now has a flexible platform for current and future use cases.

IN ONE ENGINE

Graph, vector & full-text

A Rust-native agentic graph RAG and a graph-based memory store for the agent, both on SurrealDB, instead of adding pgvector and OpenSearch to Postgres.

↑

More accurate agent responses

Graph traversal grounds the agent's responses in real relationships rather than loose vector matches.

↓

Lower token cost per call

Fetching only graph-relevant context, instead of broad vector matches, cuts tokens per call.

MONTHS

Time to production

From evaluation to a customer-facing production deployment.

MORE CASE STUDIES

See what other teams have shipped

From knowledge graphs to AI assistants - how enterprise teams are building on the context layer.

Samsung

Unlocking insights with knowledge graphs

Samsung Ads uses SurrealDB to build dynamic, real-time knowledge graphs for smarter campaign execution - collapsing three legacy data stores into one.

Read case study

Verizon

AI assistant empowering 10,000 technicians

Verizon uses SurrealDB to power a generative AI assistant for 10,000 field technicians, delivering instant access to documentation, outage updates, and workflows.

Read case study

Tencent