Devika
Open-source agentic AI software engineer that plans, researches, and writes code.
About
Devika is an open-source implementation of an agentic AI software engineer, started as a community alternative to Devin by Cognition AI. It takes high-level human instructions, breaks them down into steps, researches relevant information via web browsing, and writes code to achieve the objective, using large language models together with planning and reasoning algorithms.
The project is explicit that it remains in a very early, experimental stage with many unimplemented or broken features, and it invites contributions to move toward its stated goal of matching Devin-class scores on software engineering benchmarks. Its architecture separates an agent core, LLM backends, browser tooling, and a web UI for supervising the agent's work.
What it can do
- High-level instruction decomposition into executable steps
- Web research integrated into the coding loop
- Autonomous code writing and project scaffolding
- Supervision UI for tracking agent plans and progress
Setup
Clone the repository, install the Python dependencies, configure API keys for your LLM provider, and launch the agent server and web UI.
Prerequisites
A Python environment and API keys for an LLM provider such as OpenAI or Anthropic.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 127; Trending score 78/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.