# PuppyGraph Raises $10M Series A to Let AI Agents Query Thousands of Enterprise Tables as One Graph Without Moving the Data
> PuppyGraph raises $10M Series A led by Particle Future and defy.vc so AI agents can query up to 3,000 enterprise tables as one graph without data moves.
- **Author:** Prashant Jagtap
- **Published:** 2026-11-11
- **Updated:** 2026-11-11
- **Category:** Startups
- **Editorial type:** news
- **Topics:** ai-agents, series-a, enterprise-data, defy-vc
- **Canonical URL:** https://www.todaysstartupnews.com/news/puppygraph-raises-10m-series-a-ai-agents-graph-query-engine-enterprise-tables
## Key takeaways
- Santa Clara-based PuppyGraph has raised a $10 million Series A led by Particle Future and defy.vc, bringing total funding to $15 million
- PuppyGraph 1.0, now generally available, lets AI agents query data across up to 3,000 enterprise tables as a single graph without copying it into a separate graph database
- An AI-assisted schema builder drafts the graph schema from existing tables for teams to review, and agents connect through a built-in agent or the Model Context Protocol
- The company says annual recurring revenue grew sevenfold over the past year and names AMD, Palo Alto Networks, Datadog, Netskope, Coinbase and Trend Micro as customers
- The funding will go toward engineering and enterprise go-to-market expansion
PuppyGraph, a graph query engine startup based in Santa Clara, California, has raised $10 million in a Series A round led by Particle Future and defy.vc. The company announced the round on November 10, 2026, and says its total funding now stands at $15 million. defy.vc also led PuppyGraph's $5 million seed round in November 2024.

The money will go toward engineering and the expansion of the company's enterprise go-to-market operations. It follows the general availability of PuppyGraph 1.0, a version designed for AI agents that need to work across large volumes of company data.

PuppyGraph was founded by Weimo Liu, who is chief executive, Danfeng Xu and Lei Huang. Liu previously worked on Google's F1 team and at the graph database company TigerGraph. Xu spent nine years on LinkedIn's infrastructure team, and Huang is a three-time Google Code Jam world finalist.

The problem PuppyGraph targets is one that appears when AI agents move from tests to production. A proof of concept for an enterprise agent may cover 30 to 50 tables, but in production the same agent may need to connect up to 3,000 tables across several systems to answer a question. One example is which critical systems an attacker could reach if a particular account were compromised. As data grows, agents face more schemas, relationships and possible join paths, and complex queries can become too slow for interactive investigation or time out.

The usual answer is to load data into a dedicated graph database, which brings delays from loading, pipelines to maintain and performance limits at enterprise data volumes. PuppyGraph instead queries existing tables as one graph. According to the company, agents running on it can work across 3,000 enterprise tables and receive graph query results in under a second, without moving or duplicating the underlying data.

PuppyGraph 1.0 connects directly to platforms including Apache Iceberg, Databricks, Snowflake, ClickHouse, DuckDB and PostgreSQL. Its distributed engine splits complex graph queries into parallel operations across several compute nodes and supports petabyte-scale queries. The company says it completed a 10-hop query across billions of edges in 2.26 seconds on a cluster of four machines.

Setting up the product starts with an AI-assisted schema builder that drafts a graph schema from a company's existing tables. Data teams then review and approve it. Agents can explore the data in plain English through PuppyGraph's own agent, or through other AI tools connected with the Model Context Protocol, known as MCP. The engine rejects references to relationships that have not been defined and returns feedback, so agents can correct a query rather than guess how the data fits together. Row-level permissions apply throughout graph traversals, and single sign-on, role-based controls, user impersonation and audit logs help enterprises manage access.

Liu said the insights from enterprise data will soon be generated by agents, and that users will expect to ask questions that cross departments and systems beyond the workflows built into a dashboard. According to him, this requires agents to navigate thousands of tables, follow the right relationships and retrieve evidence in subseconds.

PuppyGraph says its customers include AMD, Palo Alto Networks, Datadog, Netskope, Coinbase and Trend Micro. The company reports that its annual recurring revenue grew sevenfold over the past year, although it did not disclose revenue figures. Ran Xia, engineering director at Palo Alto Networks, said that for a user-facing application, query latency and data freshness are essential, and that PuppyGraph lets his team run multi-hop graph traversals on its existing data platform without moving or duplicating a row. An engineer at Coinbase said that, because the exchange connects directly to tables it already keeps in Snowflake and Databricks, setting up a new graph is a schema definition rather than a data migration.

Amy Yin, partner at defy.vc, said big data remains an unsolved problem in AI. She said a user can ask an AI model to write a query, but running it across billions of relationships takes serious engineering, and that PuppyGraph was built from the start for performance at scale.

Developers can download a free version with a single Docker command and without a sign-up form, and the company says it takes about 10 minutes to go from deployment to a first query.

PuppyGraph competes with established graph database vendors such as Neo4j, which was valued at more than $2 billion after a $325 million Series F in 2021, and TigerGraph. Both are built around storing data in their own graph databases. Fortune Business Insights estimates that the graph database market was worth $2.85 billion in 2025 and will reach $20.29 billion by 2034.
## Sources
- [puppygraph.com](https://www.puppygraph.com)