# Potpie > Potpie embeds engineers in your org, maps how your systems actually work into a living context graph, and runs AI agents on it inside your own VPC. AI-native SDLC automation for large-scale engineering. Last built 2026-10-15. ## What Potpie is Potpie is AI-native SDLC automation for large-scale engineering organisations. Rather than selling seats on another tool, Potpie embeds forward-deployed engineers with one of your teams, captures how your systems actually work into a living context graph, and runs AI agents on top of that graph inside your own VPC. The company is open source at its core, is based in San Francisco and Bangalore, and is part of the Google for Startups Accelerator India class of 2026. ## How it works 1. Connect your engineering stack. GitHub, GitLab, Jira, Linear, Confluence, PLM and infrastructure, connected through approved APIs. Forward-deployed engineers embed from week one and write custom connectors and parsers for your stack. It runs inside your VPC, read-only by default. 2. Map how everything relates. Services to files, tickets to decisions, engineers to what they own. The relationships form across systems and are rebuilt as the code changes. 3. Specialised agents execute tasks on that graph: code review, debugging, spec generation, onboarding, documentation and the rest of the work that needs to know where a change fits. Potpie proves the new way of working with one team first, then helps you replicate it across the organisation. ## The context graph Code graph: AST, call graph, data flow and dependencies. Knowledge graph: entities, relations and semantics drawn from docs, PLM, wikis, standards, SOPs and design docs. Ontology layer: your domain concepts, their mapping and their relationships. Tickets and history: bugs, features, tasks and requirements, tied back to the code and the decisions that produced them. ## Models Model-agnostic by design: Claude, GPT, Gemini, xAI, DeepSeek, Llama or your own weights, through your existing subscriptions or ZDR-compliant deployments. Models can be swapped without touching workflows. Every run is logged with agent, action, tokens and outcome, so AI cost can be read against engineering value. ## Security and compliance Deployed in your VPC or on-prem. No customer data is used to train models. SOC 2, ISO 27001, ISO 42001 and GDPR. The trust centre is at https://trust.potpie.ai. The core is open source and self-hostable: your security team reads the source instead of taking anyone's word for it, and the graph and workflows are yours to keep. ## Who uses it Engineering teams at large organisations, including Fortune 500 engineering groups, use Potpie for context engineering, automated debugging and spec generation. The open source project has 5.7k+ stars on GitHub. ## Writing - [Blog](https://potpie.ai/blog): every article - [What Does It Actually Mean for an AI Agent to Understand Your Task?](https://potpie.ai/blog/what-does-it-actually-mean-for-an-ai-agent-to-understand-your-task): Engineering, Bharath Kumar, 2026-07-02 - [Why Your AI Coding Agent Needs Its Own Sandbox](https://potpie.ai/blog/why-your-ai-coding-agent-needs-its-own-sandbox): Engineering, Shambhavi Shinde, 2026-07-02 - [Evaluating AI Coding Agents in the Real World](https://potpie.ai/blog/evaluating-ai-coding-agents-in-the-real-world): Engineering, Kshitij Mishra, 2026-05-25 - [Compliance Is Architecture: Why Your AI Agent Needs a Context Graph](https://potpie.ai/blog/compliance-is-architecture-why-your-ai-agent-needs-a-context-graph): Compliance, Yash Krishnan, 2026-04-08 - [Kiro vs Potpie](https://potpie.ai/blog/kiro-vs-potpie-who-does-spec-driven-development-better): Product, Potpie Team, 2026-04-08 - [Compliance as Architecture: Building Trustworthy AI from the Ground Up](https://potpie.ai/blog/compliance-as-architecture-building-trustworthy-ai-from-the-ground-up): Compliance, Yash Krishnan, 2026-03-29 - [Building a Distributed Knowledge Graph Pipeline](https://potpie.ai/blog/building-a-distributed-knowledge-graph-pipeline): Engineering, Dhiren Mathur, 2026-03-17 - [Repository Intelligence: Building the Next Generation of Agent Evaluation Data](https://potpie.ai/blog/the-agent-evaluation-gap): Engineering, Deeptendu, 2026-03-15 - [AI Memory is Broken: we need brains, not hard drives](https://potpie.ai/blog/ai-memory-is-broken-we-need-brains-not-hard-drives): Engineering, Deepesh Genani, 2026-02-26 - [Behind the scenes : Potpie SWE Bench Agent](https://potpie.ai/blog/behind-the-scenes-potpie-swe-bench-agent): Engineering, Nandan, 2026-02-03 - [How Customer Onboarding Works at Potpie](https://potpie.ai/blog/how-onboarding-works-at-potpie): Company, Kshitij Mishra, 2026-02-01 - [Agentic COSS 2025](https://potpie.ai/blog/agentic-coss-2025): Company, Potpie Team, 2026-01-28 - [Leveraging Codebase Knowledge Graphs for Agentic Code Generation](https://potpie.ai/blog/leveraging-codebase-knowledge-graphs-for-agentic-code-generation): Engineering, Dhiren Mathur, 2025-11-14 - [Potpie supports Multi-LLMs](https://potpie.ai/blog/potpie-supports-multi-llms): Product, Dhiren Mathur, 2025-03-19 - [Release of New web access feature](https://potpie.ai/blog/release-of-new-web-access-feature): Product, Ayush Thakur, 2025-03-19 - [Open-source custom AI agents](https://potpie.ai/blog/open-source-custom-ai-agents): Product, Aditi Kothari, 2025-02-26 - [Deconstructing Potpie’s Debugging agent](https://potpie.ai/blog/deconstructing-potpies-debugging-agent): Engineering, Potpie Team, 2025-02-19 ## Links - [Home](https://potpie.ai/): the product, end to end - [Blog](https://potpie.ai/blog): every article - [Docs](https://docs.potpie.ai/): product documentation - [GitHub](https://github.com/potpie-ai/potpie): the open source core - [Trust centre](https://trust.potpie.ai/): security posture and certifications - [Careers](https://careers.potpie.ai/): open roles - [Book a demo](https://cal.com/team/potpie/potpie-s-product-demo): talk to the team