Google's New Jules Tools Is Very Cool - How I'm Using It And Other Gemini Ai Clis

Google / Elyse Betters Picaro / ZDNET

Follow ZDNET: Add america arsenic a preferred source on Google.

ZDNET's cardinal takeaways

• Google now has 3 AI CLIs for different workflows.
• Gemini CLI handles fast, interactive coding successful nan terminal.
• Jules Tools brings unreality autonomy into command-line control.

Last week, Google announced Jules Tools, a command-line companion for Google's GitHub-centric coding agent. Wait, what?

Doesn't Google already person a command-line AI coding tool? No, wait. It already has two command-line AI coding tools. Jules Tools makes for three.

Why, Google? Why must you wounded our brains so?

Before I proceed my tirade against Google's evident deficiency of merchandise management, I will show you that Jules Tools' capacity is cool. It's conscionable nan naming and merchandise fragmenting that wounded my psyche arsenic a erstwhile merchandise manager.

In July, Google announced Gemini CLI, which is precisely what it sounds like: Gemini, disposable from a terminal window. This allows for scripting, speedy interactions, and each nan different terminal-based activities we coders emotion truthful much. Then, successful August, Google announced a mouthful called Gemini CLI GitHub Actions. This provides immoderate agentic features for managing GitHub issues.

Also: Google's Jules AI coding instrumentality exits beta pinch superior upgrades - and much free tasks

Now, Jules Tools fundamentally enables Jules successful nan terminal arsenic well. Jules, arsenic you whitethorn callback from my erstwhile articles, is Google's powerful agentic coding agent, which I introduced you to backmost successful May (which, successful AI terms, seems for illustration decades agone now).

Jules Tools is simply a bully thing. But to understand wherever it fits and erstwhile you'll want to usage it, we request to deconstruct Google's CLI (command-line interface) offerings. We'll first commencement pinch Gemini, and past move to Jules.

What is Gemini CLI?

Gemini CLI is simply a command-line interface for Google's Gemini ample connection models. The CLI itself is unfastened source. What this capacity intends is that nan command-line interpreter, bid handling, nan extensions mechanism, and configuration parsing, are each disposable connected GitHub for open-source programmers to extend.

The intelligence itself comes from Gemini API calls. Gemini itself (the existent models and API backend) is not unfastened source. However, Gemini CLI is simply a instrumentality that programmers tin customize, fork, and make their own.

Also: Google embeds AI agents heavy into its information stack - here's what they tin do for you

Gemini CLI supports what are called ReAct loops, meaning it reasons and past acts. Rinse. Wash. Repeat. This characteristic intends programmers tin interact pinch it dynamically. It useful its measurement done solutions pinch nan thief of your prompts, for illustration immoderate different chatbot would, isolated from that it does truthful successful nan terminal alternatively of nan chatbot's web interface.

Operating successful nan terminal app intends you tin activity wrong immoderate VS Code-derived improvement environment. The attack useful good capable successful nan IDE to beryllium capable to know, via an IDE extension, which files are open, understand nan discourse of what you're moving on, and fto you jump backmost and distant betwixt nan IDE itself and nan terminal content.

Gemini CLI GitHub Actions runs Gemini CLI commands wrong GitHub's automation environment. GitHub, nan cardinal codification repository for galore of nan world's programming projects, has immoderate automation capabilities. Gemini CLI GitHub Actions lets you ligament Gemini into workflows, truthful it tin reappraisal code, remark connected propulsion requests, and automate different repo tasks.

What is Jules Tools?

That brings america to nan newest instrumentality successful Google's AI coding kit bag, Jules Tools. Jules is Google's cloud-based autonomous, asynchronous coding agent. Given an assignment, it develops a scheme of action and past executes that action successful its ain cloud-based virtual machine.

Also: Google's Jules AI coding supplier built a caller characteristic I could really vessel - while I made coffee

Jules Tools provides a terminal interface to Jules. You're still moving Jules successful nan cloud, but alternatively of managing it done nan Jules web interface, you tin do truthful utilizing nan CLI. The existent tasks Jules does aren't overmuch different betwixt nan 2 interfaces, but your activity style and nan imaginable for imaginative workflow optimization summation considerably.

You tin springiness Jules commands, specified arsenic create a caller task, database tasks, propulsion patches to a section instance, and more.

Some juicy examples

In Google's blog post announcing nan capability, nan authors gave 2 examples that caught my attraction and sewage maine thinking. Here's nan first example:

This codification automatically grabs nan astir precocious assigned GitHub rumor and launches a Jules distant convention to woody pinch said issue. Here's really it breaks down:

1. It uses nan GitHub CLI (gh) to database nan issues available, past filters nan issues assigned to @me, limits nan results to conscionable 1 issue, and returns conscionable nan title section successful JSON.
2. That is past piped into a JSON preprocessor (jq) to extract nan title successful plain text.
3. Then that is piped to Jules arsenic a distant task named for nan rumor title.

Whew! You tin spot really you mightiness beryllium capable to automate a batch of activity utilizing nan command-line shell, wherever nan web interface couldn't do immoderate of that.

Let's do 1 much example. This 1 provides an absorbing operation of Gemini CLI and Jules Tools successful nan bid line:

This illustration is very cool. It uses Gemini to find nan astir tedious rumor presently successful your repo.

Now, nan remark uses nan connection "difficult," but nan existent punctual uses nan connection "tedious." When asked, Gemini defines, "Tedious tasks are repetitive and boring, requiring small thought but overmuch attraction to detail. Difficult tasks are mentally challenging and require imaginative problem-solving."

Also: AI lifts immoderate package stocks, leaves others down - who's winning and losing and why

In this context, Gemini will separate betwixt difficult and annoyingly repetitive and delegate nan astir annoying and repetitive 1 to Jules. It pipes nan nonstop rumor arsenic described successful nan repo to Jules and assigns it to Jules for work.

You could virtually create a bid successful your bid statement called 'do-the-next-tedious-one,' and erstwhile you tally it, Gemini will prime it, and Jules will hole it.

Astounding.

What should you use?

Keep successful mind that it's ever champion to usage nan astir due instrumentality for a fixed task. In our workshop, we person different powerfulness saws for different types of tasks. Likewise, successful our improvement environment, we person different AI devices for different types of tasks.

Remember that some Gemini and Jules person quotas. You mightiness beryllium capable to get a batch of small tasks done successful nan Gemini tier you're using, but possibly only 15 prompts successful nan Jules tier.

Also: How AI coding agents could destruct unfastened root software

Also, Gemini responds astir instantly. Jules takes its clip to create a scheme of action, gets your support connected that scheme (in my experience, only sometimes), and past executes that plan. So, you'll beryllium overmuch much interactive and collaborative pinch Gemini, but you tin delegate to Jules.

Jules does grip large tasks, conceivably overmuch bigger than Gemini. But some are constrained by their assets allocation model. This besides deed maine when I was utilizing OpenAI's Codex successful $20/mo Plus mode. If you springiness either instrumentality a task that's excessively big, it could get blocked partially done if you tally retired of allocation. Cleanup from that business is not ideal.

I person 3 circumstantial recommendations for these tools:

1. Keep each these devices successful your kit bag.
2. Experiment pinch them erstwhile not connected a mission-critical task to get a consciousness for what they tin do and for their limits.
3. Then prime nan instrumentality (or operation of tools) that champion fits immoderate you're moving on.

What astir you? Have you tried retired Gemini CLI, Jules, aliases nan caller Jules Tools yet? Do you ideate utilizing Gemini for quick, interactive tasks and Jules for bigger, inheritance projects? How do you consciousness astir Google splitting these features crossed aggregate devices alternatively of 1 unified product? Which attack do you deliberation champion fits your workflow? Let america cognize successful nan comments below.

You tin travel my day-to-day task updates connected societal media. Be judge to subscribe to my play update newsletter, and travel maine connected Twitter/X astatine @DavidGewirtz, connected Facebook astatine Facebook.com/DavidGewirtz, connected Instagram astatine Instagram.com/DavidGewirtz, connected Bluesky astatine @DavidGewirtz.com, and connected YouTube astatine YouTube.com/DavidGewirtzTV.