Rillet's 2-person team ships 3× faster with eve agents on Vercel

Rillet sells an AI-native ERP platform whose agents do accounting work inside a real-time general ledger, with human approval and a full audit trail, and it serves more than 600 customers working toward a zero-day close. That product promise puts pressure on the company's own responsiveness: if customers are closing their books in near real time, they expect changes to arrive quickly too. Anthony Liang had been building AI applications since the GPT-3.5 era and had evaluated nearly every agent framework on the market, from the major model providers' SDKs to the open-source ecosystem. Each one gave him primitives for the model loop but left him to decide where instructions lived, how tools registered, and how an agent reached Slack and GitHub — choices he could make for his own agents but not enforce as conventions for everyone else.

He adopted eve in part because an agent becomes a directory, with instructions and skills in markdown and tools in TypeScript files whose names become the API. That structure is what let agent building spread beyond engineering, and it also sits alongside Rillet's other setup choices: Vercel Connect supplies short-lived tokens at runtime so long-lived provider secrets stay out of each application, and Enterprise Managed Users moved builders from personal accounts to company-managed ones under Rillet's identity provider, with Directory Sync adding and removing access based on assignments there. Anthony's team built a Slack bot that picks up tickets and opens PRs, while others run agents for incident analysis and Linear board management.

The broader arc is Rillet applying the same speed focus to internal tooling that it markets to finance teams. Anthony is now building what Rillet thinks of as its agent factory, so bigger features can follow the same pipeline as small fixes, with one agent writing PRDs and technical specs and another connecting to Linear boards, Miro diagrams, and the knowledge base to supply context. The headline result cited here is a two-person team closing 800+ Linear tickets in its first 4 months, and it reflects a company that has made agent conventions, credential handling, and access management part of how the team ships.