explaingit

stitionai/devika

19,504PythonAudience · developerComplexity · 4/5QuietLicenseSetup · moderate

TLDR

Open-source AI software engineer that takes plain-English instructions and autonomously plans, researches, and writes working code across multiple languages.

Mindmap

mindmap
  root((Devika))
    What it does
      Takes English instructions
      Writes complete code
      Plans and researches
      Manages projects
    How it works
      Large language models
      Web browsing
      Code generation
      Progress tracking
    Tech stack
      Python backend
      Node.js frontend
      Multiple LLM providers
    Use cases
      Build features autonomously
      Fix bugs end-to-end
      Prototype quickly
      Research and implement

Things people build with this

USE CASE 1

Tell Devika to build a new feature in plain English and let it handle planning, research, and implementation end-to-end.

USE CASE 2

Use it to quickly prototype ideas by describing what you want and letting the AI generate working code.

USE CASE 3

Have it autonomously fix bugs or refactor code by describing the problem and desired outcome.

USE CASE 4

Offload repetitive coding tasks while you focus on higher-level design and architecture decisions.

Tech stack

PythonNode.jsBunClaude 3GPT-4GeminiMistralOllama

Getting it running

Difficulty · moderate Time to first run · 30min

Requires API key for Claude/GPT-4/Gemini or local Ollama setup; Python and Node.js environments needed.

Use freely for any purpose including commercial, as long as you keep the copyright notice.

In plain English

Devika is an open-source AI software engineer that can take plain-English instructions and turn them into working code. Think of it like having an AI assistant you can tell "build me a feature that does X", and it will plan the steps, search the web for relevant information, write the code, and manage the project for you. It was created as an open-source alternative to Devin, a commercial AI developer tool by Cognition AI. Under the hood, Devika uses large language models, including Claude 3, GPT-4, Gemini, Mistral, Groq, and locally-run models via Ollama, to reason through problems and generate code across multiple programming languages. It can browse the web to gather context, extract relevant information, and track its own progress through a visual dashboard. You interact with it through a chat interface in your browser. You would use Devika if you want an AI that can autonomously handle coding tasks with minimal hand-holding, not just complete a single line, but plan, research, and write an entire feature or fix from scratch. It runs as a local server (Python backend, Node.js/Bun frontend) and requires API keys for your chosen AI model and search engine. Note that it is still in early experimental development, so expect rough edges.

Copy-paste prompts

Prompt 1
Set up Devika locally with my Claude API key and use it to build a REST API endpoint that fetches and caches data from an external service.
Prompt 2
I want to use Devika to convert a design mockup into a working React component. How do I describe the design in the chat interface?
Prompt 3
Configure Devika to use my local Ollama model instead of a cloud LLM, and have it write a Python script that processes CSV files.
Prompt 4
Use Devika to autonomously implement a feature: 'Create a dark mode toggle that persists user preference to localStorage'.
Prompt 5
How do I integrate Devika into my CI/CD pipeline so it can automatically fix failing tests and suggest code improvements?
Open on GitHub → Explain another repo

Generated 2026-05-18 · Model: sonnet-4-6 · Verify against the repo before relying on details.