Lexa AI

OpenAI variant of Encephalon Lab. Same MCP-aware agent shell, GPT-4-class models behind the chat. Built specifically to compare whether MCP agent UX feels different across model providers.

By Aditya Singh Khichi, Full Stack Engineer, New Delhi, India.

Tech stack: Next.js, TypeScript, LangChain, OpenAI, MCP, Prisma.

agent A/B: Cross-provider

Problem

Two big questions when picking a model for an MCP agent platform: does the model's native tool-calling matter more than the surrounding orchestration, and does swapping providers break user behavior in subtle ways? You can't answer either by reading benchmarks — you have to run the same UX on both and compare.

Approach

Forked Encephalon Lab's codebase, swapped Gemini for OpenAI in the agent layer, kept everything else identical: same MCP registry, same chat surface, same Prisma schema, same SSE streaming pattern. The point of the fork is exactly this — change one variable, hold everything else constant. The same MCP servers (Alpaca, GitHub, Slack, Notion) work in both, which confirms MCP's promise of model-agnostic tooling holds in practice.

Outcome

A controlled cross-provider comparison rather than a separate product. The valuable bit isn't the second platform itself — it's having both running side-by-side, which let me see where each provider's tool-calling behavior diverges. The takeaway: differences are smaller than benchmark papers suggest, and orchestration choices matter more than provider choice for end-user UX.

Live link: https://github.com/Raghav-45/lexa-ai