eve

What problem does it solve? Building backend AI agents often involves scattered configuration and opaque runtime behavior. This Skill guides you through the eve framework, where an agent is a directory on disk containing instructions, skills, tools, connections, channels, subagents, and schedules as plain files, and helps you debug deployed agents using Vercel Agent Runs observability. ## Core Features & Use Cases - Filesystem-First Agent Development: Create and edit eve agents where every component (instructions, tools, connections, channels, sandboxes, subagents, schedules, evals) is a file compiled and run by eve. - Version-Matched Documentation: Reads the bundled docs at node_modules/eve/docs/ so guidance always matches the installed eve version. - Runtime Observability: Debug deployed agents on Vercel through Agent Runs, inspecting run metadata, lifecycle events, traces, tool calls, and token usage via the Vercel MCP server or the vercel agent-runs CLI. - Use Case: You scaffold a new agent with npx eve init, define its tools and subagents as files, deploy to Vercel, then inspect a failing production run's full trace to find the exact tool call that errored. ## Quick Start Ask the assistant to scaffold a new eve agent and walk you through the bundled documentation before writing any agent code.