Creator OS
ai social agentclaude codekokonut uimcpreact components

Kokonut UI MCP Components for Your AI Social Agent (2026)

Kokonut UI exposes 100+ MIT-licensed React components over MCP, so Claude Code or Cursor pulls the right UI automatically. Learn how to wire it into a social media a

Creator OS · November 17, 2026 · 9 min read

Kokonut UI MCP components give an AI social agent a library of React building blocks it can pull in over MCP, so you spend prompts on what matters and skip writing every button by hand. It is an open source set of components for chat and product UIs. Pair it with the tool that actually runs your social accounts, and you can see a working dashboard sooner.


  your prompt
      |
      v
  Claude Code / Cursor
      |            \
      |             \ MCP
      |              v
      |        Kokonut UI MCP  --> components
      v
  Creator OS MCP  --> posts, comments, analytics
      |
      v
  social scheduling UI you run locally

Two MCP servers, two jobs: one ships UI, one moves data.

What is Kokonut UI and why does it matter for a social media agent

Kokonut UI is an MIT licensed library of 100+ React components served over MCP, and that matters because your agent can request a component on demand instead of you pasting one in.

The problem with most component libraries is they are a catalog you browse. You find the card, copy the source, adjust the import, fix the theme. That is fine for a marketing page. It is slow for a tool you are building with an agent in a loop.

MCP changes the shape of the work. The library becomes a server your agent talks to. You ask for a reply queue, Claude Code or Cursor fetches the components it needs, and you review the result. The components are the raw material. You still decide layout, state, where the data comes from and what happens on a click.

Get started for $0

How the Kokonut UI MCP server works with Claude Code and Cursor

The pattern is the same for both: register the server once, then ask for a component by name or by description.

MCP servers speak a small set of tools to the AI app. A UI component server exposes “what do you have” and “give me that one”. Your agent sends a request for, say, a stat card or a search input, the server returns the component source plus props, and the agent writes it into your project with the imports already correct.

Claude Code and Cursor both support MCP servers. You add the entry to your config, restart, and the agent sees new tools in its context.

Where you work What you do What the agent gets
Claude Code Register the server with the CLI or settings Component tools plus your repo
Cursor Add an mcpServers entry Component tools plus your open files
Any MCP app Same server, same tools Component tools plus whatever else you connect

The one thing Kokonut UI does not do is know your social data. It has no idea what a comment is or how many followers you gained. That is the job of the social layer. See the MCP docs for what Creator OS exposes.

Kokonut UI vs hand-writing components for a social agent dashboard

The fastest path is not “use a library” versus “write everything”. It is a mix: prebuilt UI for the ordinary pieces, your own code for the parts that decide.

Hand writing gives you full control and full cost. Every table, badge, empty state and modal is an hour. A component MCP server gives you a starting point you can change. The tradeoff is you still own the data wiring, the loading states and the edge cases.

Piece Better from a library Better written by you
Stat cards, tabs, inputs Yes Rarely worth it
Reply queue rows Layout and styling Send, hide, reply logic
Analytics panel Charts and containers Which metric, from which call
Ad report table Table and filters Spend and budget rules

Think of it as scaffolding. The agent gets the visual shell in one prompt, then you point it at real calls and the shell becomes a tool.

How to wire Kokonut UI MCP into your social media agent workflow

Wire both servers into the same agent, then let one write the UI and the other supply the data contract.

Do it in this order so you are never guessing:

  1. Connect the UI server so your agent can pull components.
  2. Connect Creator OS so the agent can read live posts, comments and analytics.
  3. Ask the agent to build the smallest useful screen first: a reply queue.
  4. Run it, click everything, then widen.

A prompt that works as a starting point:

Pull the list components and the stat card components from
Kokonut UI. Build a reply queue page. Each row: platform,
author, comment text, a Reply button and a Hide button.
Load rows from the Creator OS MCP tool list_comments with
platform instagram. Reply calls reply_to_comment, hide calls
hide_comment. Show a toast on success. No mock data.

Two things make this work. The agent knows the tool names from the connected server, so it does not invent an endpoint. And the UI components arrive ready to render, so it spends its reasoning on the flow instead of a div soup.

The same loop works for scheduling. Ask for a compose form plus a preview panel, and have the submit call create_post. Validate captions before you send them with the tools covered in the social media API for AI agents.

What the MCP component library gives your agent: reply queues, analytics panels, ad reports

The value shows up as whole screens, each one a small set of components plus a few real calls.

A reply queue is a list, a textarea and two actions. The data comes from live comments. On Creator OS the reply tools cover Instagram, Facebook, X, Threads, YouTube, LinkedIn and Reddit. If you build a Hide button, it maps to the same networks: the hide tool backs comment hiding on Instagram, Facebook, X, Threads, YouTube, LinkedIn and Reddit. Your agent can read the rows, show them, and send the answer back.

An analytics panel is stat cards on top of one query. Creator OS exposes per post analytics, daily metrics, follower stats and best time to post, so the agent can fill a chart without you writing a transform. It can also read Reel watch numbers like skip rate and average watch time after a post goes out, then suggest a tighter hook on the next video. Creator OS provides the posting and the numbers. Your agent’s prompt decides what to change.

An ad report is a table over the Ads add on. If you run ads, the agent can list ads, read spend and results, and pause what is not working. Build the table from components, then wire it to the ad tools. Ads are created paused for review, and a dry run is available with validate_only, so a stray click does not spend money.

Can you build a social scheduling UI without writing components by hand

You can get most of the way there, and the part you still write is the part that matters: the decisions.

Here is a realistic split of a first working version:

  • From the MCP library: the shell, navigation, cards, forms, tables, badges, empty states.
  • From your agent: the page wiring, which tool each button calls, loading and error states.
  • From you: the rules. What counts as a good reply. When to approve. What to never automate.

That last column is not small. A scheduling UI where a click publishes to eleven platforms is a sharp tool. Creator OS marks destructive actions so your AI app asks first before anything is deleted or disconnected, and the Claude Code plugin adds a publish guard that asks Proceed or Cancel before anything is published, spent or deleted. Keep guardrails in the UI, not just in the prompt.

If you want to watch the build happen end to end, the walkthroughs on the KevBuildsApps YouTube channel show the agent loop in practice, and the launch video covers the tooling around it.

Social media MCP for Claude: where Kokonut UI ends and Creator OS begins

Kokonut UI makes the interface. Creator OS makes the accounts move. Neither one tries to be the other.

Creator OS is one place to publish, schedule, reply and read analytics on 11 platforms plus Skool, a WordPress blog, a Shopify store and paid ads. It ships a hosted MCP server at https://mcp.creatoros.ca/mcp with about 160 tools, and nothing to install. Claude connects through Settings then Connectors, ChatGPT through a connector in developer mode, and Claude Code through one command:

claude mcp add --transport http creatoros \
  https://mcp.creatoros.ca/mcp \
  --header "Authorization: Bearer cos_live_..."

Read only connections only see read tools, and the API refuses writes from them. Agency plans get one scoped key per client workspace, so an agent holding one client’s key cannot post to another.

The division looks like this:

Layer Kokonut UI MCP Creator OS MCP
Job Ship React components Run your social accounts
Output Component source and props Posts, comments, analytics, ads
Data None Live accounts and metrics
Good for Dashboards, forms, tables Doing the work the UI triggers

Use both in one session and you get a dashboard wired to real accounts without a month of front end work. The Claude Code build guide walks the agent side, and the skills guide covers the packaged workflows that ship with the CLI.

FAQ

What are Kokonut UI MCP components?

They are React components from the Kokonut UI library, exposed through an MCP server so an AI app can request them by name. Your agent pulls the component it needs instead of you copying source out of a catalog. Kokonut UI is MIT licensed and lists over 100 components.

How do I add the Kokonut UI MCP server to Claude Code?

Register it as a stdio or HTTP MCP server in your Claude Code configuration, matching however the project tells you to run it, then restart so the tools appear. Confirm it loaded by asking the agent what components it can see. For a hosted server, the command shape is the same as any remote server: claude mcp add with a transport and a URL.

Can an AI social agent build its own dashboard UI?

It can build the interface when it has a component source and a data source, and it can assemble a working reply queue or analytics page in a session. It cannot invent your accounts, so connect a social layer with MCP tools that read and write real posts. The rules about what to automate should stay with you.

Is Kokonut UI free and MIT licensed?

Yes, Kokonut UI is open source under the MIT license, so you can use the components in commercial work. Check the repository for the current component count and setup steps before you build. Creator OS is a separate paid product and does not include Kokonut UI.

What is the best social media MCP for Claude?

Pick the server that can do the work you need: publish, reply, read analytics and manage ads across your accounts. Creator OS ships a hosted MCP server with about 160 tools covering those, plus a CLI and a Claude Code plugin, and it is included in every plan. Start from the docs.

How does Creator OS use MCP components for social media agents?

Creator OS does not ship UI components. It ships the tools your UI calls, plus skills that arrive as MCP prompts like meta-campaign-builder, so they appear in your AI app’s prompt menu. Pair it with a component server and your agent has both the look and the actions.

Get started with Creator OS

Connect the component server for the interface, connect Creator OS for the accounts, and let your agent build the first screen today.

Creator OS plans start at $14.99/month for 11 platforms plus Skool, with the API, the MCP server and the CLI included. New subscribers start for $0 with a free trial and can cancel anytime. Sign up here, then read the agents docs and the authentication docs before you hand a key to anything.

Get started for $0
Free trial for new subscribers. Cancel anytime.

Keep reading