Industry: Artificial Intelligence (AI)
Outcome: unified agent memory with multi-tenancy and real-time sync
DigitalKin (https://www.digitalkin.com) is a French startup building an agentic-mesh ecosystem to revolutionise workplace efficiency by automating repetitive digital work. Its Kins-autonomous digital employees - integrate into teams, boosting productivity so humans can focus on meaningful interactions. To power memory, cross-agent coordination, and live UI state, the team needed a single, responsive data layer. After moving to SurrealDB DigitalKin unified document and relational access, enabled push updates via live queries, and isolated tenants with namespaces, with graph relations next to enrich memory and relationships.
DigitalKin’s customers operate autonomous agents with strict data boundaries. The team needed tenant isolation that didn’t force a separate database instance for every organisation, while keeping operations straightforward across environments.
Polling created laggy dashboards and brittle webhook chains as agents learned, acted, and exchanged tools. The platform required push-driven updates so interfaces and agents stayed aligned with authoritative state.
The team needed schema-optional modelling to iterate quickly, allowing their customers to develop agents that can structure and store data freely, while operating within defined constraints.
DigitalKin was looking to implement native similarity search and relationship modelling to represent agent-to-agent and agent-to-tool links. The stack had to start with document and relational models, then adopt graph traversal without re-architecting.
SurrealDB lets Digitalkin provide turnkey graph and vector storage for agents, enabling native RAG and KAG. Beyond search and indexing, it maps relationships between agents and tools across the agentic mesh so each agent can quickly route to the most relevant resource. The stack evolved from document/relational models to add graph traversal and vector search without a major re-architecture.
SurrealDB’s flexible schemas let DigitalKin deliver “a database within the database”, provisioning a turnkey collections and records model so agent developers can structure and store data freely, while operating within the defined constraints and indexes.
Live queries stream changes to the platform as they happen, replacing polling and glue code, and keeping both the agent mesh and enterprise dashboards in sync.
Each organization has its own namespace for clean data isolation and access control - no separate clusters needed. Production runs on Surreal Cloud for managed backups and easy scaling, while a self-hosted node is used for local iteration and automated tests; a consistent API across environments prevents drift and speeds releases.
Shared infrastructure with namespace isolation reduced per-tenant overhead while preserving strict boundaries.
Surrealist accelerated query design, schema evolution, and debugging, improving developer velocity for SDK and protocol work.
Live queries enabled push-driven UX and inter-agent events, keeping interfaces and agents aligned without polling.
The team is preparing native vector indexing and graph relations to consolidate agent memory and knowledge with relationships.
DigitalKin chose SurrealDB for its hybrid data model that unifies relational joins and document-style deep fetching, plus live queries that simplify real-time state. Namespaces delivers tenant isolation without per-customer clusters, and the API-native interface - available in both managed Surreal Cloud and self-hosted deployments - keeps development and production aligned. With Surrealist and SurrealQL, the team iterates quickly today and has a clear path to add vectors and graph tomorrow, all within one operational engine.
“SurrealDB is more than just our database, it’s the backbone of our Agentic Mesh ecosystem. Its speed, flexibility, and multimodal capabilities empower us to unify and scale the complex data flows that fuel our AI agents, both now and in the future.” - Thibaud Perrin, DigitalKin CTO
Get in touch to learn more →