Skip to content
Kodel.AI Docs

Introduction

Kodel.AI is an enterprise coding agent that integrates directly into your software development workflow. It connects your project management, git repositories, and AI models to automate repetitive engineering tasks.

How it works

Section titled “How it works”
Browser ──► Server (:3000) ──► LLM (OpenAI / Anthropic / ...)
  SSE ◄       │ ▲
              │ │ WebSocket
              ▼ │
            Agent (Docker)
            /workspace
  1. You create a task in the web UI (or via API)
  2. The server assigns it to a running agent over WebSocket
  3. The agent uses an LLM to write code, run tests, and fix errors
  4. Results stream back to your browser in real time via SSE
  5. The agent opens a pull request when done

Key features

Section titled “Key features”
  • Multi-tenant — each organization has isolated data
  • Multi-LLM — switch between providers per project
  • Streaming — watch the agent work in real time
  • Tool use — the agent reads files, runs bash, searches code
  • Audit log — every action is recorded

Architecture overview

Section titled “Architecture overview”

Kodel.AI is a client-server monorepo built with:

LayerTechnology
ServerNestJS + PostgreSQL + ClickHouse
AgentDocker container + Bun runtime
Web UIReact + TanStack Router + Tailwind
ProtocolWebSocket (agent ↔ server) + SSE (server ↔ browser)

Next steps

Section titled “Next steps”