Cosmic AI Agents vs. Contentful Skills: Two Very Different Visions of AI in the CMS

Contentful ships Contentful Skills, a free, MIT-licensed package that teaches AI coding agents how to build on Contentful. It works with Claude Code, Cursor, GitHub Copilot, VS Code, OpenAI Codex, Gemini CLI, and other platforms that support the agentskills.io spec. It is genuinely well-built and solves a real problem for Contentful developers.

Credit where it is due. What follows is an explanation of why Contentful Skills and Cosmic AI Agents solve different problems, and why that difference matters when you are choosing where to put your content stack.

Everything below about Contentful Skills was verified against Contentful's own repository on the date in the verification line at the end of this post. Where the repository does not establish a claim, we say so instead of guessing.

What Contentful Skills Actually Does

Contentful Skills ships as seven skills:

• contentful-guide: Core concepts and routing. Explains spaces, environments, content types, entries, assets, and locales, then points you at the right skill.
• contentful-api: Language-agnostic API reference with curl examples covering their Content Management, Content Delivery, Content Preview, and Images APIs.
• contentful-nextjs: SDK setup, environment variables, content fetching, Draft Mode previews, and ISR for both App Router and Pages Router.
• contentful-migration: Writes and runs content model migration scripts, including field operations, validations, entry transforms, and editor interface wiring.
• contentful-custom-app-from-scratch: Designs, scaffolds, builds, and validates new App Framework custom apps.
• contentful-custom-app-enhancement: Improves and debugs existing custom apps in customer-owned repositories.
• contentful-personalization: Sets up, debugs, and develops personalization and A/B testing with their Experiences SDK.

The personalization skill is the most involved. It runs three workflows (Onboard, Doctor, and Develop) backed by a reference library the repository describes as 19+ documents, built on a structured state machine with MCP server integration. The Doctor workflow diagnoses broken setups by checking packages, environment variables, API connectivity, and content state. That is real engineering, not a docs wrapper.

Installing the plugin also registers two MCP servers: a hosted Contentful MCP server for CMS operations, described in the repository as reading and writing content, content types, and assets, plus a local MCP server that powers the personalization workflows.

The repository states the goal in one line: "Your AI agent knows how to code. These skills teach it Contentful."

That is the product, and it is a good one. Contentful Skills makes developers faster and more accurate when they build on Contentful. When you ask your coding agent a Contentful question, it now has context-aware, codebase-aware answers instead of generic ones.

What Cosmic AI Agents Actually Do

Cosmic takes a different shape. There is one kind of agent, and you decide what it can do by enabling capabilities on it. A single agent can hold several at once, which means the same agent can research a topic, write the post, commit the code change, and report back in Slack.

The capabilities you can enable:

Team communication

Agents live in Slack, WhatsApp, and Telegram with their own persona, persistent memory, and goals. A team member can send a message like "Write a blog post about our Q2 product update and save it as a draft" and the agent does it. Agents also support scheduled check-ins, so they can surface information proactively rather than only when asked.

Content

Create, update, and bulk-manage content objects. Create object types and metafields. Generate images and video. Publish automatically or route everything through human review.

Code

Connect an agent to a GitHub repository and it works on isolated branches, commits, opens pull requests, and deploys to preview or production. A human reviews the PR.

Browser automation

Drive a real browser: authenticated sessions, form fills, visual data extraction, and recorded demo videos.

Supporting capabilities

Agent delegation so one agent can hand work to another, external API requests, encrypted secrets, and runtime log analysis.

Goals and Workflows

Recurring work is expressed as goals. A goal is an instruction the agent works through in a loop until it judges itself done, and a goal with a schedule repeats on that cadence. For a fixed sequence of steps, Workflows chain agents together with scheduled, webhook, or manual triggers, and steps can run in parallel. If you want to build one, the lesson on building an AI agent team that ships content walks through it.

There is also an Agent Marketplace of pre-built agents and workflows you can install in one click rather than configuring from scratch.

The Difference That Matters: Scope of Action

This is where the two products diverge.

Contentful Skills improves the developer. The developer remains the actor. They prompt their coding agent, the skill supplies better guidance and can reach into Contentful through the MCP server, and the developer ships better code faster. This is the copilot model, and it is a strong implementation of it.

Cosmic agents are the actor. A marketing manager can message an agent in Slack and get a draft post. A content ops team can schedule an agent to audit product descriptions nightly. An agent connected to a repository can pick up an issue, write the fix, open a PR, and post the link to Slack while nobody is watching.

Which one fits depends on the problem in front of you.

If your problem is that your developers lose time to documentation and setup on Contentful, Contentful Skills is well-targeted and free. Install it.

If your problem is that content creation, publishing, deployment, and distribution all require a human to start each step, you are looking for agents that operate on their own.

Side-by-Side Comparison

One note on how to read this table. The Contentful Skills column describes what is in that package, verified from its repository. It is not a claim about everything Contentful the platform can do, and rows marked as outside the package should be read that way.

What This Looks Like in Practice

A content lead messages an agent in Slack: "We need a post comparing headless CMS options for Next.js teams, targeting 'best headless CMS Next.js'. Save it as a draft and send me a summary." The agent researches, writes with proper structure, generates a featured image, saves the draft, and posts a summary back with a review link.

An engineering team sets a weekly goal: every Monday, check open GitHub issues tagged priority:high, write fixes on isolated branches, open PRs with descriptions, and post the links to the engineering channel.

A marketing team runs a Workflow: a webhook fires when a feature ships, one step drafts the announcement, another posts a teaser to Slack, another records a short demo video for social. Every step is logged.

Where They Overlap: MCP

Both products have an MCP server, and both let an AI coding tool reach into the CMS. Cosmic's MCP server gives Cursor, Claude Code, and other MCP clients real access to your content. That overlap is genuine, and if MCP access is all you need, both platforms give it to you.

The overlap is a slice of the Cosmic AI story. The larger part is agents that keep working while you are doing something else.

Pricing and Access

Contentful Skills is free and MIT-licensed. Install it and your coding agent gets Contentful context. For teams already on Contentful, that is a good deal with no downside.

Cosmic agents are included on every plan, including Free:

• Free: $0/month, 1 agent
• Builder: $49/month, 3 agents
• Team: $299/month, 10 agents
• Business: $499/month, 25 agents
• Enterprise: custom

Workflows are a separate pool at the same counts per plan, and workflow steps do not count against your agent limit. Only active agents count toward it. Plans include a set number of team members (Free 2, Builder 3, Team 5, Business 10), and additional users are $29/user/month.

The Free plan needs no credit card, so you can build a working agent today without spending anything.

The Bottom Line

Contentful Skills is a thoughtful piece of developer tooling. If your team is on Contentful and your developers work inside AI coding tools, install it. It will save them real time, and the MCP server it registers gives their agent genuine read and write access to your content.

Cosmic agents answer a different question: what happens when the work itself runs without a person starting it. Writing, publishing, deploying, distributing, and reporting back, on a schedule or a trigger, with humans reviewing output rather than producing it.

If you are deciding where AI fits in your content stack in 2026, that distinction is the one worth being clear about before you commit to a platform.

---

Ready to see what autonomous agents look like in practice? Start building free on Cosmic, no credit card required. Or book a 30-minute demo with Tony to see a full agent workflow live.

---

Last verified: August 25, 2026. Contentful Skills details verified against github.com/contentful/skills. Cosmic plan and agent counts verified against cosmicjs.com/pricing. Contentful's own pricing page serves a JavaScript verification challenge to automated clients, so this post makes no claim about Contentful plan pricing.

---

Related Reading

• What the Salesforce Acquisition Means for Contentful Customers (2026): Why the Salesforce acquisition is accelerating the evaluation cycle for Contentful customers, and what Cosmic offers as an independent alternative.
• The Best Contentful Alternative in 2026: A full breakdown of why teams are choosing Cosmic as their Contentful alternative, with pricing comparisons and feature details.
• How to Migrate from Contentful to Cosmic in 30 Minutes: A step-by-step technical guide to migrate from Contentful to Cosmic with TypeScript SDK code for every step.
• How Enterprise Teams Are Replacing Contentful with an AI-Native CMS: What large organizations look for when evaluating a Contentful replacement, including security, SLAs, migration support, and AI capabilities.