MonkeysCode has been introduced—a new agentic IDE with an open-core architecture that prioritizes privacy and complete user control over the development process. The platform allows for flexible switching between proprietary models, such as Claude, Gemini, and ChatGPT, and local solutions via Ollama or vLLM, while ensuring an unprecedented level of data security.
What Happened
Developers have unveiled MonkeysCode, an agentic development environment that supports on-device code indexing and the ability to work in a fully isolated (air-gapped) mode. The system includes mechanisms for verifying agent actions using signed logs and requires mandatory test passes before any code changes are applied.
Context
Unlike popular closed-source solutions such as Cursor or Claude Code, MonkeysCode utilizes an open-core model. This allows developers not only to choose any LLM, including fully local ones, but also to obtain a transparent audit of the AI agent's work, which is critical for ensuring project integrity.
Why It Matters for the Industry
The emergence of tools capable of operating in air-gapped environments and providing audit mechanisms for agent actions lowers the barriers to adopting AI coding in regulated sectors, such as fintech and the public sector. This paves the way for using agentic AI in companies with strict security and code traceability requirements.
Why It Matters for Users
For developers, this means the ability to use any set of models (including Llama 3 or Mistral) without vendor lock-in. The automated testing system serves as a protective barrier, ensuring that the agent does not introduce destructive changes that fail the project's existing checks.
Sources
Author
Look at AI, Editorial Team