Potpie AI raises USD $2.2m to map massive codebases

Potpie AI has raised USD $2.2 million in a pre-seed round as it builds a context layer designed to help AI agents work inside large, complex software codebases.

Emergent Ventures led the round, with participation from All In Capital, DeVC, and Point One Capital. Potpie plans to use the funding for early enterprise deployments, engineering hires, and continued work on its context and agent infrastructure.

Engineering teams have adopted generative AI tools quickly, but many products focus on code generation. Large language models often struggle without system-level context, tooling history, and architectural intent. In large organisations, that context is spread across internal systems and processes-and often concentrated in the heads of a small number of senior engineers.

Potpie positions its software as a way to unify context across the engineering stack. It describes the approach as "spec-driven development" and says the product ingests information from source code repositories, ticketing systems, logs, documentation, and code reviews.

Context layer

Potpie says its platform makes the specification the source of truth. Agents first turn requirements into an implementation plan, then map dependencies and edge cases and align tests and rollout steps before any code is written.

The company also describes its approach as "ontology-first," building a graphical representation of software systems and inferring behaviour and patterns across modules. It says the platform produces structured artefacts that guide agent behaviour within a given codebase.

"As AI makes code generation easier, the real challenge shifts to reasoning across massive, interconnected systems. Potpie is our answer to that shift, an ontology-first layer that helps enterprises truly understand and manage their software," said Aditi Kothari, CEO and co-founder of Potpie.

Potpie targets large enterprises with codebases starting at around one million lines of code, and says it can scale to hundreds of millions. It lists use cases across the software development lifecycle, including debugging cross-service failures, maintaining and writing end-to-end tests, blast-radius detection, and system design.

System updates

Potpie says the product updates context as systems evolve. It can update documentation and tickets when pull requests are created, and generate system designs when new tickets are opened.

The company says it generates structured definitions for each agent, describing how the agent should operate within a specific codebase. It also builds a searchable index across APIs, services, databases, and components, using tags to narrow results across a system.

Founders and traction

Kothari founded Potpie with Dhiren Mathur. The pair began working on the product in October 2023, according to the company. Potpie says the founders spent nearly two years building the underlying layer that understands codebases and creates a knowledge graph, ahead of a public launch in January 2025.

Potpie says early deployments highlight the scale of the problem in large organisations. One customer with a codebase of more than 40 million lines reduced root-cause analysis for production issues from nearly a week to around 30 minutes, according to the company. Engineers reportedly shifted from running investigations to reviewing output.

The company also cited a second customer with decades-old systems. That customer used Potpie to update and generate tests in the background, cutting work that previously took multiple sprints into a shorter cycle.

Investor view

Anupam Rastogi, managing partner at Emergent Ventures, said: "In large enterprises, the real challenge is not generating code; it is understanding the system deeply enough to change it safely. Potpie's ontology-first architecture, combined with rigorous context curation and spec-driven development, creates a structured model of the entire engineering ecosystem. This allows AI agents to reason across services, dependencies, tickets, and production signals with the clarity of a senior engineer. That is what makes Potpie uniquely capable of solving complex RCA, impact analysis, and high-risk feature work even in codebases exceeding 50 million lines."

Potpie says it works with Fortune 500 and publicly listed companies in regulated industries, including healthcare and insurtech. It also says its open-source projects have more than 5,000 stars on GitHub.

"AI readiness is not about picking the right model," said Kothari. "It's about building systems that can support intelligence over time. Our goal is to make Potpie the foundational layer engineering teams rely on to build, operate, and evolve complex software with AI built in from the start."