PuppyGraph raises $10 million to give enterprise agents a structured route through company data
PuppyGraph raises $10 million to give enterprise agents a structured route through company data
PuppyGraph announced a $10 million Series A on November 10, bringing its total funding to $15 million as it expands a graph-query engine for enterprise AI agents. Particle Future and defy.vc participated. The company's announcement positions the software as a way to investigate relationships across business data without first copying that data into a separate graph database. PuppyGraph announcement
The round follows general availability of PuppyGraph 1.0. SiliconANGLE reports that the company plans to expand engineering and its enterprise sales operation. SiliconANGLE
Making relationships explicit
PuppyGraph's approach starts with a model of business entities and the relationships between them. An agent can query those defined connections rather than guessing which database columns should be joined. The company says an undefined relationship produces structured feedback that the agent can use to revise its query. The underlying records remain in existing enterprise systems. PuppyGraph announcement
SiliconANGLE describes an AI-assisted schema setup process that data teams approve before use, plus support for connecting other agents through the Model Context Protocol. Its reporting also describes short-lived, user-linked credentials for Snowflake and Databricks connections and enforcement of existing row-level permissions. Those capabilities are reported product behavior, not results from an independent security audit. SiliconANGLE
The quality of the map still matters
Guth News analysis: Rejecting an undefined relationship can prevent one category of invented query, but a defined relationship can still be wrong or incomplete. The organization remains responsible for deciding what its business entities mean, how access rules apply and whether an agent's conclusion follows from the returned records.
For prospective customers, a useful evaluation would therefore test deliberately ambiguous questions and incomplete schemas, not only the speed of a successful demonstration. Teams should also check whether evidence and permissions survive a multi-step investigation. Funding and general availability make this a concrete deployment option; they do not establish that every generated business explanation will be reliable.
Sources rechecked November 15, 2026. Both named investors participated; sources differ on which led the round.
Sources and citations
Revision history
Published .