Public Roadmap for Gemini CLI v1

[ryanjsalva]

Well, butter my biscuits! 🍞 The last two weeks have been hella exciting for Gemini CLI. If you haven’t seen the stats, check this out:

• 58.8k stars on the repo which has grown over 10x since our launch day
• 710 issues opened/week and 312 closed/week
• 320 PRs opened/week and 120 merged/week
• 12 releases which means we’ve released almost every weekday since Gemini CLI was first announced

We’re moving fast, responding to issues (thank you for reporting them 🙏), and have a clear vision for the future – one that we hope inspires you to build, learn, and contribute. To create a shared vision for the product, today we’re excited to announce the Gemini CLI public roadmap and project board.

The public roadmap is designed to answer: what new capabilities will ship in the coming months, and when should you expect them? With more transparency into what we’re building, you can also plan better and influence our direction. Especially over the next 3 months, we're focused on building a command line tool that is:

1. Extensible: Gemini CLI should be adaptable to a wide variety of use cases and environments through emerging extensibility like Remote Model Context Protocol (MCP), custom /slash commands, hierarchical GEMINI.md files, etc.
2. Everywhere: Developers should be able to invoke background agents, allowing you to run Gemini CLI in containers anywhere (e.g. locally, via GitHub Actions, via Jules, and in-the-cloud) configured by you. Gemini CLI should also be able to delegate tasks to sub-agents running in those containers.
3. Intelligent: Gemini CLI should be reliably ranked among the best agentic tools as measured by benchmarks like SWE Bench, Terminal Bench, and CSAT. To meet these benchmarks, we also aspire to fix all P0 defects before declaring v1.
4. Free & open source: We want to foster a thriving open source community where cost isn’t a barrier to personal use, and PRs get merged quickly. This means resolving and closing issues, pull requests, and discussion posts quickly. Our ambition is to maintain a sustained count of <200 open issues and <200 open PRs – though we may adjust those numbers as the community participation rate also increases.

🙋 How do I contribute?

Our sincere hope is that, by providing transparency into both our roadmap and design principles, we make it easier for everyone to contribute.

Before writing code, we ask that everyone familiarize yourself with our contributor guidelines and open an issue for discussion. Pull Requests (PRs) attached to an issue and aligned with the Gemini CLI roadmap will receive first priority. PRs that don’t align with the roadmap are likely to be rejected (sorry). PRs addressing reliability issues, product defects, UX quirks, and documentation gaps are always appreciated, but (you guessed it) should also reference an issue.

💬 Keep the feedback coming!

Your feedback is crucial to improving Gemini CLI for everyone. And if I'm being honest, your stories are what inspire many of our core contributors to dedicate their time to this project. So please, share your stories and super-user tips for using Gemini CLI. We all want to learn from you.

As always, happy coding! 👾

[erinmikailstaples]

Thank you so much for this visibility and transparency! Rooting for y'all!

[OGODEVO]

Loving the cli

[AlexsandroAF]

This is great news about the Gemini CLI! The rapid growth in stars, the active issue and PR handling, and the frequent releases are all really impressive. I especially appreciate the commitment to transparency with the public roadmap and project board. Outlining the focus on extensibility, ubiquity, intelligence, and the dedication to open source is very helpful for the community. The call for contributions is clear, and emphasizing the importance of aligning PRs with the roadmap and referencing issues is a good way to manage contributions effectively. Thanks for sharing the update and the vision for the future of the Gemini CLI! I'm looking forward to seeing how it evolves. I will be sure to submit any feedback and share my experiences as I use the CLI.

[matiasmolinas]

Hi @ryanjsalva and the Gemini CLI team,

First, thank you for sharing this detailed public roadmap! The transparency is incredibly valuable for the community, and the vision you've laid out for v1 is inspiring. The focus on making the CLI more Extensible, Intelligent, and capable of running agents Everywhere is exactly what the ecosystem needs to build truly powerful, next-generation workflows.

I'm writing to respectfully draw your attention to a pull request I've submitted, as I believe it provides a direct and robust implementation for several of the core pillars you've outlined.

PR #3174: feat(core): virtual tools support (#1806)

This PR introduces a "Manifest-Driven Virtual Tool System" that directly aligns with your roadmap in the following ways:

1. Extensibility: It provides a concrete and powerful evolution of the "hierarchical GEMINI.md files" concept. It transforms GEMINI.md from a static context file into a fully executable manifest, allowing users to define custom, sandboxed tools using simple shell scripts. This dramatically lowers the barrier for creating project-specific capabilities, a key goal of making the CLI more adaptable.

2. Everywhere: The roadmap mentions the goal to "delegate tasks to sub-agents." My PR includes a reference implementation of this concept via a run_agent virtual tool in my LLMunix testbed project. This tool demonstrates how the Gemini CLI can be used to invoke isolated sub-processes, a foundational pattern for building complex, multi-agent systems that can run anywhere.

3. Intelligence: To excel at benchmarks like SWE Bench, agents need resilience and the ability to self-improve. The architecture enabled by this PR fosters these advanced agentic behaviors. By allowing an agent to define and call its own tools, it can dynamically evolve its capabilities to solve complex problems, as demonstrated in the PR's detailed test plan.

The feature is designed to be 100% generic, opt-in, and secure by default, as it leverages the CLI's existing sandboxing infrastructure. My goal was to provide a high-leverage, low-surface-area change that aligns perfectly with the future direction of the project.

I have included a comprehensive test plan in the PR description, including an end-to-end acceptance test using the open-source LLMunix project, which serves as a real-world validation of this architecture's power.

I'm very excited about the future of Gemini CLI and would be grateful if the team could take a look at PR #3174 when time permits. I believe it's a significant step towards achieving the v1 vision.

Thank you again for your incredible work and for fostering such a vibrant open-source community!

[yayxs]

is really great, looking forward to a better Gemini CLI

[marcelo-schroeder]

Thank for you the transparency. Gemini CLI is already great and it can become even better.

[aniketpotabatti]

Love using Gemini cli ❤️ excited for more cool updates.

[CostcoFanboy]

I feel giving Google One/Google AI Pro/Ultra subscribers extra credits or a shared pool of usage like Anthropic does would push adoption rates for both products through the roof.

Would there be any plans for that? I don't see anything regarding this in the roadmap.

[murasz]

Are you planning to make a plugin like Cline/Roo Code inside the IDEs in the future or will the interface always stay CLI?