The Pulse
PuppyGraph raises $10 million for enterprise AI agent data tools
Particle Future led PuppyGraph’s Series A, bringing the company’s total funding to $15 million. The company plans to expand engineering and enterprise go-to-market work for its graph query engine.
PuppyGraph has raised $10 million in a Series A round to expand its graph query engine for enterprise AI agents. In a November 10 post, CEO and co-founder Weimo Liu said Particle Future led the round, with participation from defy.vc, bringing the company’s total funding to $15 million.
The company plans to use the money for engineering and enterprise go-to-market expansion. Its priorities include reducing the work needed to establish a graph model, supporting relationship queries at enterprise scale and helping more teams put agents into production.
Querying relationships in existing data
PuppyGraph 1.0 runs graph queries over supported databases, warehouses and data lakes. The underlying records stay in the enterprise’s existing data systems; teams do not need to load them into a dedicated graph database or maintain a separate graph copy and loading pipelines.
The company says the software already enables agents to query across 3,000 enterprise tables as one unified graph. Its approach is intended to let agents discover relationships and retrieve evidence as an investigation develops, rather than putting every table definition into a single prompt.
The product uses an ontology—a model of business entities, such as customers, deals and products, and their relationships. Its AI-assisted schema builder examines existing tables and proposes entities and relationships; the data team reviews and approves the model, including business meaning that table structures alone cannot establish.
Checks on agent queries
If an agent refers to an entity or relationship that the ontology does not define, PuppyGraph rejects the reference and returns structured feedback. The agent can then inspect the model and correct its query, potentially finding a path through other entities or determining that the relationship has not been modeled.
Liu cautioned that the model does not guarantee a correct conclusion: an agent may still form a poor hypothesis or conduct an incomplete analysis. The check is aimed at preventing an agent from treating an invented relationship outside the approved model as evidence.
PuppyGraph supports MCP, which lets compatible AI tools and its built-in agent access enterprise data through graph queries. Supported systems include Apache Iceberg, Databricks, Snowflake, ClickHouse, DuckDB, PostgreSQL, Microsoft Fabric and Cloudera.
Applications can also use an existing SQL query engine for filters and aggregations and turn to graph queries to follow relationships between entities. Version 1.0 enforces configured row-level permissions during graph traversals, including intermediate steps. Role-based access controls and user impersonation tie data access to user identity.
Performance and customers
PuppyGraph says it supports subsecond graph queries for supported workloads. In a separate company benchmark, it completed a 10-hop query across billions of edges in 2.26 seconds on a four-machine cluster. The post notes that performance depends on the workload.
The company lists AMD, Palo Alto Networks, Datadog, Netskope, Coinbase, Trend Micro, Baker Hughes and Sola Security among its customers. Annual recurring revenue grew 7x over the past year, according to PuppyGraph.
Liu said AMD began with IT workloads and that this work has opened the door to connecting information security and digital marketing data. The company also described customers expanding from observability into fraud detection and from initial agent workloads into additional business domains.