Connect Claude to Apatero and Generate AI Images From a Conversation

I spend half my working day inside Claude, and until recently every image meant leaving it. Draft in the chat, jump to a generator, upload references, download results, drag them back into the conversation. You can now connect Claude to Apatero through a single connector URL and skip that entire loop, because images generate and render inside the chat itself.

This guide covers the whole setup end to end. You get both connection paths, claude.ai and Claude Code, the first prompt to try, where results actually render on each surface, what every generation costs in credits, and fixes for the connection problems people actually hit.

Quick Answer. In claude.ai, open Settings, then Connectors, then Add custom connector, and paste https://mcp.apatero.ai/mcp. Sign in with your Apatero account when the OAuth window appears, then ask Claude for an image in plain language. The result renders directly in the conversation, and a default image generation costs 9 credits.

What The Apatero Connector Actually Does

MCP stands for the Model Context Protocol, an open standard for connecting AI assistants to outside tools and services. Claude speaks it natively, which means any MCP server can hand Claude new abilities without waiting for an official integration. Apatero runs one of those servers, and it turns Claude into the conversational front end of a full AI creative studio.

Once connected, Claude sees 20 tools it can call on your behalf. I wrote a separate breakdown of all 20 tools if you want the complete catalog, but these are the ones you will actually use in your first week.

• apatero_generate_image handles text to image and can also edit with reference images you provide
• apatero_generate_video runs Seedance 2.0 for text to video or image to video
• apatero_talking_avatar uses Kling Avatar v2 Pro to speak a script through TTS or lip sync to audio you supply
• apatero_create_influencer generates a portrait set and saves the character as a reusable Soul
• apatero_generate_with_soul puts that same face into any new scene you describe
• apatero_instadump turns one photo into a batch of 3 to 15 Instagram ready posts
• apatero_show_image redisplays any earlier image when you want another look
• apatero_credit_balance, apatero_list_models, and apatero_remix handle balance checks, model discovery, and remixing

The important part is that you never call these by name. You describe what you want in normal language, and Claude picks the right tool on its own.

Connect On Claude.ai In One Minute

The claude.ai path uses a custom connector, and Anthropic supports those on every plan. Free, Pro, Max, Team, and Enterprise accounts can all add one, so this works even if you have never paid for Claude. Anthropic keeps an official guide to custom connectors using remote MCP if you want the background reading, but the short version follows.

You also need an Apatero account, since generations run against its credit system. Signing up is free and takes about as long as the connector setup itself.

1. Sign in at claude.ai
2. Open Settings, then Connectors
3. Click Add custom connector
4. Paste https://mcp.apatero.ai/mcp as the connector URL and confirm
5. When the sign in window appears, log in with your Apatero account to complete the OAuth step
6. Start a new conversation and ask for an image

That is genuinely the whole procedure. The OAuth step links your Apatero account to Claude, so every generation you run from chat spends credits from that account, which keeps billing in one predictable place.

Connect In Claude Code

If you live in the terminal, Claude Code connects to the same server with one command. The difference is authentication, because instead of an OAuth window you pass an API key in a header.

Create the key first. In the Apatero app, open the Developer section at app.apatero.ai/developer and generate a key there. Then run this in your terminal.

claude mcp add --transport http apatero https://mcp.apatero.ai/mcp --header "Authorization: Bearer YOUR_APATERO_API_KEY"

Swap in your real key and you are done. Claude Code now has the same 20 tools as claude.ai, which makes it easy to fold creative work into development sessions. I have asked for placeholder art in the middle of a frontend task and kept coding while it landed, and that pattern alone justified the setup for me.

Your First Image From A Conversation

You do not need any special syntax. Ask for an image the way you would ask a colleague, and Claude calls apatero_generate_image with a prompt it writes from your request. The default model is Nano Banana 2, which costs 9 credits per image.

Here are a few starter prompts that show the range.

• "Generate an image of a cozy reading nook at golden hour, photorealistic, warm tones"
• "Make a flat vector style illustration of a rocket launch for a slide deck, white background"
• "Here are two photos of my product. Generate a lifestyle shot that keeps the product accurate"

Iteration happens in the same thread. Tell Claude the sky should be darker or the framing should be wider, and it runs an edit pass with the earlier image as reference instead of starting from scratch. If a result scrolled out of view, ask for it again and apatero_show_image redisplays any earlier image. You can also ask what models exist or what your balance looks like, and Claude answers through apatero_list_models and apatero_credit_balance without you touching a dashboard.

All of this works whichever Claude model you happen to run. If you are still deciding which model fits your daily work, my Claude Haiku 4.5 complete guide covers the speed and cost tradeoffs in detail.

What Renders Where

The part that surprises people the first time is that results do not arrive as bare links. On claude.ai and Claude Desktop, images render in an inline Apatero panel built on the MCP Apps extension, sitting directly in the conversation flow. In Claude Code, images render inline in the terminal conversation, so even a text interface shows you the actual picture.

Video and talking avatars behave a little differently because they take longer. Those renders run roughly 1 to 5 minutes, so the tool returns a generation id immediately and Claude checks the status until the result is ready. You ask for the video, and Claude reports back with the finished render instead of leaving you to refresh anything.

What Generations Cost

Generations spend credits on the connected Apatero account, not on your Claude subscription. Your Claude plan only determines whether you can add connectors at all, and since custom connectors work on every tier, the real variable is your Apatero credit balance. Current plans live on the Apatero pricing page, and you can check your remaining balance mid conversation whenever you want.

One more cost note before you plan a big session. Souls, InstaDump, and the create influencer flow involve uploads, so they need an active plan on the Apatero side. A default image at 9 credits remains the simplest and cheapest way to confirm the connection works.

A Full Workflow From Blank Chat To Finished Assets

Here is the exact flow I use for blog artwork, because it shows how the pieces stack together. Everything happens in one conversation, and you can ask for your balance along the way to watch the running total.

1. Describe the hero image and let the default Nano Banana 2 model take the first pass for 9 credits
2. Give feedback in plain language and run an edit pass against the first result for another 9 credits
3. Upscale the winner with SeedVR for 2 more credits before downloading

That is a finished, polished hero image for 20 credits without ever leaving the chat. The edit pass is where the conversational format earns its keep, because you refine with sentences instead of rewriting prompts from zero.

Character work goes one layer deeper. The create influencer tool generates a portrait set and saves the character as a reusable Soul, and from then on apatero_generate_with_soul can place the same face in any new scene you describe. I wrote a full walkthrough on how to create an AI influencer inside Claude if consistent characters are your goal.

When you need volume instead of a single hero shot, InstaDump takes one photo and returns a batch of 3 to 15 Instagram ready posts. That tool exists for feeding a posting schedule from a single source image, and it runs from the same conversation as everything else.

Troubleshooting Connection Problems

Most failed setups trace back to a small set of causes. Work through these in order and you will almost always land on the fix within a couple of minutes.

The Connector Will Not Save On Claude.ai

Check the URL first. It must be exactly https://mcp.apatero.ai/mcp, including the /mcp path at the end. A pasted URL that lost its path or picked up a trailing space is the most common failure I have seen, and retyping it cleanly usually resolves the add step.

The OAuth Sign In Fails Or Loops

The sign in window wants an Apatero account, not your Claude login. If you do not have one yet, create a free account before adding the connector, then retry the add flow and complete the sign in when the window reappears.

Claude Code Rejects The Connection

In Claude Code the header does the authenticating, so a placeholder or expired key fails. Confirm you replaced YOUR_APATERO_API_KEY with a real key from the Developer section at app.apatero.ai/developer, then rerun the add command with the corrected header. Generating a fresh key and repeating the command is faster than debugging an old one.

Claude Describes Images Instead Of Generating Them

If Claude writes a paragraph about your image rather than making it, verify the connector shows as added under Settings, then Connectors. After that, start a fresh conversation and ask directly for a generated image, since an explicit request removes any ambiguity about whether you wanted a description.

Generations Fail Partway Through

Ask for your credit balance right in the chat. Every generation spends credits from the connected Apatero account, so an empty balance stops image calls even when the connection itself is healthy. Plans are listed on the pricing page if you need to move up a tier.

Souls Or InstaDump Return Errors

These features involve uploads and need an active plan on the connected account. If plain image generation works but character tools or InstaDump fail, the account plan is the first thing to check before you suspect the connector.

A Video Render Seems Stuck

Video and avatar renders take roughly 1 to 5 minutes, which feels long next to an image. The tool returns a generation id immediately, so ask Claude to check the status of that generation instead of submitting the job again and spending a second round of credits.

Key Takeaways

If you skimmed to this point, here is the entire guide compressed into six lines.

• The connector URL is https://mcp.apatero.ai/mcp, and it works on every claude.ai plan from Free through Enterprise
• Claude Code connects with one claude mcp add command plus an API key from app.apatero.ai/developer
• Images render inside the conversation, in an inline Apatero panel on claude.ai and Claude Desktop and as inline terminal images in Claude Code
• Default images cost 9 credits with Nano Banana 2, upscales cost 2, and Seedance 2.0 video runs 165
• Credits spend from the connected Apatero account, and Souls, InstaDump, and influencer creation need an active plan
• Video and avatar renders take 1 to 5 minutes, and Claude tracks the generation id so you never babysit a queue

Frequently Asked Questions

Does The Apatero Connector Work On The Free Claude Plan?

Yes. Custom connectors are available on Free, Pro, Max, Team, and Enterprise plans on claude.ai, so everything in this guide works without a paid Claude subscription. Your only costs are the Apatero credits each generation spends.

Do I Need An Apatero Account Before Connecting?

You do, because the OAuth step signs you into Apatero and every generation spends credits from that account. Signup is free, and the connector can be added the moment the account exists.

How Much Does A Single Image Cost?

The default model is Nano Banana 2 at 9 credits per image, and its edit mode is also 9 credits. GPT Image 2 runs 40 credits for either generation or editing, and a SeedVR upscale adds just 2 credits on top of any image.

Can Claude Generate Video Through This Connector?

Yes. The apatero_generate_video tool runs Seedance 2.0 for text to video or image to video at 165 credits, renders take roughly 1 to 5 minutes, and Claude checks the generation status for you until the result comes back.

Where Do The Generated Images Show Up?

Directly in the conversation. On claude.ai and Claude Desktop they render in an inline Apatero panel built on the MCP Apps extension, and in Claude Code they appear as inline images in the terminal itself.

What Is MCP And Do I Need To Understand It?

MCP is the Model Context Protocol, an open standard for connecting AI assistants to external tools. You do not need to understand the protocol to use the connector, the same way you do not need to understand HTTP to read a website.

Can I Reuse The Same Character Across Different Images?

Yes, that is exactly what Souls are for. The create influencer tool generates a portrait set and saves the character as a reusable Soul, and apatero_generate_with_soul then renders that same face in any new scene you describe.

How Do I Check My Credit Balance Without Leaving Claude?

Just ask. The apatero_credit_balance tool reports your remaining credits inside the conversation, which is worth doing before you queue a 165 credit video render.

The whole setup comes down to one URL on claude.ai or one command in Claude Code. Create a free Apatero account, add the connector, and ask Claude for your first image. The first render costs 9 credits, and the tab switching loop never comes back.