jules-skill: Control Google's AI Coding Agent Jules from the Command Line

When AI Coding Assistants Meet the Command Line

Google's AI coding agent Jules is becoming a new addition to developers' toolboxes, and the ecosystem of tools built around it is starting to take shape. The open-source project GreyC/jules-skill on GitHub is one such effort — it provides a CLI skill that lets developers manage and invoke Jules directly through a REST API.

The project is still small in scale (3 stars, 1 fork, primarily written in JavaScript), but it addresses a genuine pain point: how to seamlessly integrate an AI coding agent into existing terminal workflows, rather than being locked into a specific GUI or web application.

A Multi-Platform Design Philosophy

The most noteworthy aspect of jules-skill is its compatibility positioning. According to the project description, it doesn't serve a single environment — it simultaneously supports Claude Code, Gemini CLI, and standard terminal use cases.

This design reflects a broader trend in the AI coding tool ecosystem: developers often use multiple AI assistants at once and need a unified way to orchestrate them. By wrapping Jules as a reusable CLI skill, jules-skill lets users invoke Jules's capabilities directly within Anthropic's Claude Code or Google's own Gemini CLI, forming a cross-tool collaboration pipeline.

For engineers who prefer to do most of their work in the terminal, this "everything via command line" approach typically means greater efficiency — no more switching back and forth between browsers and IDEs, and AI agent calls can be embedded directly into scripts and automation workflows.

REST API: The Key to Connectivity

The core mechanism of the project is communicating with Jules through a REST API. This means jules-skill is essentially a lightweight API client wrapper that translates Jules's exposed interfaces into more accessible command-line instructions.

This architecture offers several practical advantages:

• Low coupling: As long as Jules's API remains stable, the tool continues to work without depending on any specific UI version
• Programmability: REST interfaces are naturally suited for scripted invocation, making integration into CI/CD pipelines or automated tasks straightforward
• Cross-language potential: While the current implementation is JavaScript-based, REST's universality means similar wrappers can be reproduced in any language

For developers who want to explore the boundaries of Jules's capabilities, interacting directly through the API is also an effective way to bypass UI limitations and probe raw functionality.

Who Is It For

jules-skill is still in its early stages, with relatively lean features and documentation. It's best suited for the following types of users:

• Developers already using Jules who want to integrate it into their command-line workflow
• Power users who work with both Claude Code and Gemini CLI and want multiple AI tools to work in concert
• Tech enthusiasts interested in API-level access to AI coding agents who are willing to tinker

It's worth being upfront: this project has limited community activity and maturity at this stage. It feels more like a proof of concept or personal utility than a production-ready solution. Users who want to try it should have solid command-line and API experience, and be prepared for rough edges along the way.

A Small Tool Reflecting a Larger Trend

Beyond the project's current scale, jules-skill reflects a broader direction in which AI coding tools are evolving from "closed applications" into "composable components." When every AI agent exposes an open API, the developer community can chain them together like building blocks, constructing personalized intelligent workflows.

Community-built adapter tools like this are often an early signal of an AI product's ecosystem health. The fact that Jules is attracting developers who write CLI skills for it shows it's already building real usage demand among engineers. As these tools proliferate, the boundaries between AI coding assistants will grow increasingly blurry — eventually merging into every terminal command a developer types on any given day.