Skip to main content
Vision & Future

Claude Code roadmap

Upcoming features and improvements planned for Claude Code: what to expect in the coming months and how to prepare.

The Claude Codex launched in early 2026. In a few weeks, the site went from a landing page to a comprehensive guide of over 50 pages covering installation, MCPs, Skills, agents, prompting, enterprise topics, and much more. Here's a recap of what's been done, what's coming, and how you can contribute.

What's been accomplished

Phase 1: Foundations (January-February 2026)

The first weeks laid the technical and editorial groundwork for the project.

Infrastructure:

  • Tech stack: Next.js 14 (App Router), TypeScript strict, Tailwind CSS, MDX
  • Static export (SSG) for maximum performance
  • Docker deployment (Nginx Alpine), image < 50 MB
  • Full SEO: metadata, JSON-LD, sitemap, canonical URLs
  • Dark mode / Light mode, mobile-first design, Lighthouse score > 90

Initial content:

  • Landing page with main feature overview
  • "Getting started" section: installation, configuration, first project
  • MCP section: understanding, installation, top MCPs by category
  • Prompting section: basics, directives, templates, common mistakes
  • Skills section: concepts, top skills, custom skill creation
  • Agents section: concepts, creation, agent teams, orchestration
  • Vision & Future: overview page with trends and roadmap

Phase 2: Deep dive (March 2026)

Content expanded to cover the needs of beginners and experts alike.

New content:

  • Beginner paths: zero prerequisites, 40+ term glossary, FAQ
  • Editorial articles: real costs, myths, security best practices
  • Enterprise section: security, adoption, TCO, FAQ, governance
  • Advanced section: hooks, headless/CI-CD, multi-provider
  • Technical reference: cheatsheet, CLI, settings.json, environment variables
  • Limits section: known limitations, Copilot and Cursor comparisons, when not to use
  • Persona paths: developer, lead, non-dev, freelance, student
  • Advanced MCP: TypeScript creation, Python creation, advanced protocol
  • Interactive configurator: config generator in a few clicks
  • Vision section: themed pages on AI and careers, 2026 trends

Technical improvements:

  • Built-in search with accent normalization
  • Inter-page navigation with previous/next
  • Auto-generated table of contents on every page
  • Breadcrumb and structured data on all sub-pages

What's coming

Here are the features and content planned for the coming months. Priorities may shift based on community feedback.

Q2 2026: Interactivity and community

Under development

These features are in preparation. Dates are indicative and may change.

Interactive tutorials

The goal is to offer exercises directly in the browser. You'll be able to write a prompt, see the simulated result, and understand why one wording works better than another, without leaving the site.

Video guides

Visual walkthroughs for the most requested topics: first installation, creating an MCP, configuring agents, multi-tool workflows. Short format (5-10 minutes), on YouTube and embedded in pages.

Community contribution system

A clear process for submitting content: recipes, workflows, experience reports. Contributions will be reviewed and integrated into the guide with credit to the author.

Q3 2026: Directory and tools

MCP & Skills catalog

A searchable directory of all available MCPs and Skills, with detailed cards: description, author, last update date, security rating, installation instructions.

Prompting playground

A space to test prompts and compare results between different wordings. Useful for learning prompting without consuming tokens.

Q4 2026: Internationalization

Multilingual support

The guide translated into English (priority 1), then Spanish, Arabic, and other languages based on demand. The goal: making content accessible beyond the French-speaking world.

How to contribute

The Claude Codex is an open project. Every contribution is welcome, from fixing a typo to writing a full chapter.

Report a problem

Spotted an error, a broken link, or outdated information?

Open an issue

Go to the GitHub repository and click "New Issue."

Describe the problem

Indicate the page concerned, what is incorrect, and if possible, what the right information should be.

We'll take care of it

The team triages issues regularly. Simple fixes are often published the same day.

Propose content

Have a workflow, recipe, or experience to share?

Fork the repository

Clone the project locally and create a branch with a descriptive name (feat/my-article).

Write your content in MDX

Place your file in content/ with complete frontmatter (title, description, section, order). Check existing files for the format.

Submit a pull request

Push your branch and open a PR. Describe what you're adding and why. The team will review and guide you if adjustments are needed.

Contribution guidelines

A few rules to help contributions fit in well:

  • Language: English, accessible and direct tone. No unnecessary jargon, no hollow phrases
  • Style: no em dashes, no AI-sounding wording ("It's important to note that...", "Dive into...")
  • Structure: clear introduction, structured body with h2/h3 headings, conclusion with "Next steps"
  • MDX: use the available components (<Callout>, <Steps>, <Card>, <CodeBlock>) where relevant
  • Frontmatter: title and description are required. Add order for navigation positioning

Need help?

If you're not comfortable with Git or MDX, simply open an issue with your content in plain text. We'll handle the formatting.

Long-term vision

The ambition of The Claude Codex is to become the go-to reference for Claude Code and, more broadly, for using AI in development.

What that means in practice:

  • Completeness: cover every topic, from absolute beginner to expert deploying in an enterprise. No gaps in the learning path.
  • Quality: every page should be useful, up to date, and well written. 50 excellent pages beat 200 mediocre ones.
  • Accessibility: the site must remain free, without a paywall or intrusive tracking. Technical knowledge should not be a privilege.
  • Community: over time, the guide should be as much community-driven as it is team-driven. The best open-source technical resources work on this model.
  • Honesty: we talk about limitations as much as strengths. No marketing, no hype. If a tool doesn't work well for a use case, we say so.

What this project is not

To be transparent about the project's intentional boundaries:

  • It's not a commercial site. No product to sell, no paid course behind a form
  • It's not Anthropic's official documentation. It's an independent guide, with an editorial point of view
  • It's not an AI news site. We don't cover every release; we focus on what's durably useful

Next steps