Agent Memory vs.
agent databases

Q: Does Agent Memory support MCP?

Yes. Agent Memory exposes seven MCP tools: remember, recall, context, reflect, forget, upload and inspect. It also has a REST API with an OpenAPI specification, and Python and TypeScript SDKs.

Q: Is the platform production-ready for agent workloads?

Yes. SurrealDB has been in development since 2018 and runs in production at Samsung, Verizon, Tencent and PolyAI. Enterprise deployments come with SOC 2 and ISO 27001 compliance, dedicated support and SLAs.

Explore Agent Memory Start in Studio

Agent databases. Each covers a part: HelixDB has no document model, ReasonDB no graph, HydraDB no shared transaction.

The gap. Two stores in one box are still two stores, with no transaction across the graph and the vectors.

SurrealDB. Graph, vector, document and structured data in one engine and one transaction.

THE LANDSCAPE

Built for agents, assembled from parts

HelixDB

A Rust graph-vector database with its own query language and MCP support. Strong developer experience; no native document model or enterprise governance.

HydraDB

Memory infrastructure with context evolution tracking. Graph and vector are separate substrates, with no transaction across them.

ReasonDB

A document database with LLM-guided tree traversal in place of flat chunks. Novel retrieval; no graph model and no temporal versioning.

THE COMPARISON

Two stores in one box are still two stores. Where the agent databases stop, and where Agent Memory carries on.

Data and transactions

Agent Memory

Agent databases

ACID across models

One transactionGraph, vector, document and time commit together, or not at all.

Two commitsThe graph store and the vector store commit separately.

Document model

Full documentsSchema-validated documents; an edge is a full document with fields of its own.

Nodes and edgesEvery record is a node or an edge; the document is somewhere else.

Operations

Governance

In the databaseRow and field-level permissions, JWT auth and tenant isolation, enforced by the engine.

In the applicationAccess control lives in the application code around the store.

Storage

SeparatedCompute apart from storage: scale to zero and built-in recovery.

CoupledCompute and storage scale together, on the same nodes.

THE ALTERNATIVE

One engine, one transaction

Transactions

ACID across every model

Graph traversal, vector similarity, document operations and temporal queries commit in one transaction.

How transactions work

Edges

Edges are documents

A relationship carries metadata, timestamps, embeddings and structured fields of its own.

How RELATE works

Governance

Governance built in

DEFINE ACCESS gives row and field-level permissions, JWT auth and tenant isolation. SOC 2 and ISO 27001 certified.

Explore permissions

HOW IT WORKS

One commit for the fact and the state. A graph write, a vector and a document change land together, or not at all.

1BEGIN TRANSACTION;
2
3-- The decision, the memory of it and the relationship, in one commit
4UPDATE order:1042 SET status = "refunded";
5
6RELATE agent:support->remembers->customer:acme SET
7    fact = "Prefers a refund to store credit",
8    embedding = fn::embed("prefers refund over credit"),
9    source = ticket:8812,
10    valid_from = time::now();
11
12COMMIT TRANSACTION;

TRUSTED BY

Enterprise teams building on SurrealDB

See more case studies

Three data stores collapsed into one

An AI assistant for 10,000 technicians

Nine backend tools consolidated into one

Low-latency RAG, under the customer's control

45 million recommendation queries a month

Agent databases,answered

How is Agent Memory different from HelixDB or HydraDB?

Agent Memory is built on SurrealDB, where documents, graphs, vectors, relational and geospatial data are native models, and facts, graph relationships, vectors and their history are written in one transaction. Agent databases run separate graph and vector substrates without a transaction across them or enterprise governance.

Does Agent Memory support MCP?

Is the platform production-ready for agent workloads?

THE PLATFORM

Everything an application and its agents know. Five surfaces, one engine.

Database

Document, graph, vector, time-series and relational in one engine.

Five data models as dotted tiles: documents, graph, vector, time-series and relational

Agent Memory

What an agent learns, with its source and its time, in the same engine.

A timeline of remembered facts, each with its source

Cloud

Dedicated clusters by region, or serverless on SurrealDB Earth.

Clusters in three regions on a world map, each running or scaling

Studio

Query, explore and design the schema from the browser.

MCP

Every model that speaks MCP reaches the database and the memory directly.

Three models connected through MCP to the database and Agent Memory

IN PRODUCTION

Trusted at scale. Samsung, Nvidia, Verizon, Tencent, and Walmart run on SurrealDB.

14,000+

Developers building on SurrealDB Cloud

4M+

Developers building on SurrealDB worldwide

FROM THE TEAMS

SurrealDB gives us a foundation where we can unify semantic search, knowledge graphs, and AI-driven decision making without stitching together multiple systems. Collapsing responsibility into SurrealDB has become our default engineering posture.

Justin Foley

VP of Engineering, Later

GET STARTED

Build on proven foundations. Both run on one engine. The database is a free account away. Agent Memory is a context in Studio away.