ECA (Editor agnostic coding assistant) (eca.dev)

ECA (Editor Code Assistant) is an open-source, editor-agnostic protocol and server that connects code editors to LLMs for a unified AI pair‑programming experience. Editors run a small plugin that spawns an ECA server (written in Clojure and inspired by the Language Server Protocol) and communicates via stdin/stdout. ECA provides a chat interface, single global/local configuration across editors, and an “agentic” mode where LLMs can act with configurable tools and MCPs. It already supports multiple providers — OpenAI, Anthropic, GitHub Copilot, Ollama local models and custom backends — and can auto-download the latest server on editor start. Technically significant features include multi‑model login and runtime selection, context enrichment (project prompts and MCP resources), tool-call management for chaining model actions, and OpenTelemetry export of usage and metric data. That server‑middle design makes it easy to add capabilities (telemetry, tool orchestration, consistent UX) without changing each editor plugin. For the AI/ML community this standardizes editor↔model integration, simplifies multi‑model experimentation and local model adoption, and surfaces telemetry that can help evaluate UX and model behavior in real projects. The project is contribution‑friendly with a public roadmap and quickstart flow (configure a model, /init to create AGENTS.md, start chatting), and is seeking maintainers and sponsors.