PuppyGraph has raised a $10 million Series A to expand its graph query engine for enterprise AI agents. Particle Future led the round, with participation from defy.vc, and the new capital takes the Santa Clara startup’s total funding to $15 million.
The round arrives alongside the general availability of PuppyGraph 1.0. The release lets AI agents query data across as many as 3,000 enterprise tables as one connected graph, without moving the underlying records into a separate graph database.
What PuppyGraph 1.0 adds
PuppyGraph 1.0 adds an AI-assisted schema builder that drafts a graph model from existing tables for a data team to review and approve. It also checks every query against that model and rejects references to entities or relationships the model does not define.
The engine runs multi-hop graph queries directly over the warehouses, lakes and databases a company already uses. It supports Apache Iceberg, Databricks, Snowflake, ClickHouse, DuckDB, PostgreSQL, Microsoft Fabric and Cloudera, and agents can reach it through MCP or the built-in agent.
Security is built into the traversal itself. PuppyGraph 1.0 enforces configured row-level permissions at every step of a query, including intermediate steps, so an agent only sees data its user is allowed to see.
Why agents need a graph
Most AI assistants that work with data rely on text-to-SQL, which works well when the tables and relationships are already understood. The harder problem starts when an agent must discover those relationships across thousands of tables on its own.
A wrong join can produce a valid looking answer built on a relationship that does not exist in the business. PuppyGraph gives the agent a defined model of customers, deals, products and suppliers to explore, so it queries relationships the business has approved instead of inventing its own.
Speed matters because an agent tests many hypotheses in a single investigation. PuppyGraph reports subsecond graph queries for supported workloads, and it completed a 10-hop query across billions of edges in 2.26 seconds on a four-machine cluster.
Early traction
PuppyGraph says its annual recurring revenue grew seven times over the past year, without disclosing the absolute figure. Its customers include AMD, Palo Alto Networks, Datadog, Netskope, Coinbase, Trend Micro, Baker Hughes and Sola Security.
“We run AI agents across thousands of tables and billions of edges a day.”
Leon Goldberg, CTO of Sola Security
defy.vc previously led the company’s $5 million seed round in November 2024. PuppyGraph will use the new funding for engineering and enterprise go-to-market expansion, and a free developer edition is available through a single Docker command.
