An agent-heavy roadmap for 2026

There’s a clear theme in our new set of proposed roadmap items: automation and agent readiness. It’s the first major update since the introduction of our new product strategy, so it feels important to get into the rationale.

Why automation

This is best explained with the obligatory XKCD webcomic: Is It Worth the Time?.

This chart is pre-ChatGPT, these days things are moving so much faster that automation becomes more of a necessity for us to keep up with the robots. And using LLMs to write code, the time investment is drastically lower. We can much more easily create deterministic, 🤖✨-free automation. So higher return, and lower investment.

Here are examples of what this will mean in Wagtail:

• Better, more advanced content checks in the CMS. Like heading density and quality, for accessibility or SEO benefits.
• Automated website carbon footprint calculations. And publication with the carbon.txt standard.
• Deep integrations and customizations with other systems. Like having access to performance reports from within the CMS to make better decisions.
• More sustainable open source maintenance, like our security vulnerability processes. Where our contributors’ time is getting stretched thin by the continued waves for AI-generated contributions of varying quality.

Wagtail’s code-first approach to content modeling has always been excellent for this. And our Python / Django foundations are excellent for that too, with their immense package ecosystem. And now, we’re all more able than ever to capitalize on this.

Why agent readiness

Let’s be clear upfront; if you want to make your site more agent-ready, start with a passing accessibility audit. The web generally, and well-designed accessible content sites are already very agent-friendly as it is. Wagtail consistently excels to help sites meet strict conformance with well-established accessibility standards (WCAG, ATAG), that were already conceived with the goal to make content work for a wide range of user agents.

If you already meet that baseline, there are emerging standards and patterns that are worth (gradually) introducing on sites. But they’re lost in a sea of hyperbole and snake oil. We’re navigating this and plan to work on the no-nonsense aspects of agent readiness and GEO (Generative Engine Optimization), with a standards-based approach. With improvements such as:

• Built-in Markdown support / machine-readable formats (✨ RSS! ✨ ).
• Improvements to content quality for extraction, like better anchor links support and modern signals of E-E-A-T (experience, expertise, authoritativeness, and trustworthiness).
• Emerging standards like Well-Known agent skills discovery and WebMCP.

This won’t be relevant for all sites – there are lots of excellent reasons to keep content out of reach of LLMs for specific projects or organizations. We’ll focus on the bits that drive content quality up for everyone as core features, and build more opinionated or experimental improvements in separate packages, per our AI guiding principles.

Where automation meets agent readiness

The two kinds of automation meet with Developer Experience with agent skills, our plan to improve how AI models big and small perform on common Wagtail tasks. The goal here is to produce a solid benchmark and identify opportunities to improve our documentation for all of our users, with a select few skills there to fill the remaining gaps. We expect most developers will want to install two to three skills for their Wagtail-Django usage, not 10.

People focusing on smaller models might want to have more ready-made skills and we’ll want to cater that too, as long as we can be confident that our work does meaningfully improve the quality or efficiency of the tasks.

From our DX with AI survey, here are the kinds of tasks we have in mind:

We’ll likely pick a mix of frequent tasks where following best practices generally helps, and rarer but critical operations where we want more constrained behavior from agents.

Beyond automation and agents

Wagtail isn’t becoming an AI project despite the agent-heavy roadmap. The focus is on producing high-quality content and code, and avoiding the slop. To figure this out, more than anything, we need your feedback! Our proposed roadmap items are all detailed in RFC 116, currently under core team review and awaiting community feedback.

Here are specific future additions which need feedback (and feature sponsors) more than the rest to move forward:

• Shareable previews (with people who do not have CMS access)
• Generic content importer - copy-paste from Word reinvented to be a one-step process creating a nicely structured page.
• GEO power tools - quality assurance tools so you know how your content looks in AI overviews.