Every AI model,
right inside your AI app
Connect Cabina via MCP and use 55+ models — GPT, Claude, Gemini, Flux, Midjourney, Runway, Veo and more — right inside Claude, ChatGPT, Cursor or any MCP client. One secure connection: text, images, video and audio without switching apps.
Your AI app
The AI tools you already use
Cabina MCP
A secure connection via the MCP protocol
Cabina inside your AI
Text, images, video, audio + tools: RAG, Inpainter, Upscaler
Connect Cabina to Claude
One click in the browser — or one command in the terminal.
Sign in to Cabina
A Cabina account is needed to confirm access in the next step.
Log inNo account yet? Create one for free
Click Connect
The button opens Claude with the form already filled in — nothing to paste.
In the window that opens, sign in to your Cabina account and click “Allow”.
Start creating
No more setup — just type it into the Claude chat:
Sign in to Cabina
A Cabina account is needed to confirm access in the next step.
Log inNo account yet? Create one for free
Add it as an MCP server (not just a link)
The URL alone isn’t enough — open your client’s MCP settings and add it as a new server (or use its add-server command).
https://api.cabina.ai/mcpIn the window that opens, sign in to your Cabina account and click “Allow”.
Start creating
No more setup — just type it into the Claude chat:
auth Authorization: Bearer YOUR_TOKEN
https://api.cabina.ai/mcpWorks with the tools you already use
Cabina speaks native MCP — drop it into any MCP-compatible client in one command. No glue code, no waiting.
One adapter between your agent and every model
MCP (Model Context Protocol) is the open standard that lets coding agents and AI clients plug straight into outside tools. Connect Cabina once and your agent reaches all 55+ models behind a single key — no SDKs to wire up, no per-vendor accounts, no copy-pasting between tabs.
Instead of leaving your editor to generate an image or render a clip, you just ask: “use cabina to make a hero image with Flux” — and the asset comes back in your workflow.
Connect in one command3 steps. One token. You're live.
No browser, no SDK wrangling. Create a key, copy the token once, and Cabina answers from your code or your agent.
Create a key & copy the token
Generate a personal API key in your profile. The token appears once at creation — store it safely.
Authorize your requests
Pass the token as a Bearer header. The same key powers both raw API calls and your MCP connection.
Generate from anywhere
Call text, image and video models from scripts, bots and backends — or expose Cabina as an MCP server to Claude Code, Codex, Kimi, Cursor and more.
Built for builders, safe by default
Everything you need to wire Cabina into your stack — and nothing in your way.
One command. You're live.
No SDK lock-in, no boilerplate, no server to host. A single CLI command and Cabina lives right inside your MCP client.
- Works in Claude, Codex, Cursor & more
- OAuth or API key — your choice
- First request in under a minute
Token-based security
Every key is a scoped Bearer token, shown only once. Rename, regenerate or revoke any of them independently — anytime.
Shown once01Works with your whole stack
Plain REST for scripts and backends, or MCP for AI agents. One key — you choose the surface.
REST + MCP02Live in under a minute
Copy, paste, ship. Keys are active the instant you create them — no review queues, no approvals.
Instant03Many keys, clean control
A separate key per environment, service or teammate. Flip Active / Inactive without touching the rest.
Granular04Automate everything
Drive Cabina from CI/CD, cron jobs and internal tools — backend-to-backend, no human in the loop.
Hands-free0555+ models across every modality
One key reaches the whole roster — and new models are added the moment they ship, at no extra cost.
Chat & reasoning
Image generation
Video generation
Voice & transcription
Just ask your agent
Plain-language prompts in your CLI — Cabina picks the model and returns the result inline.
Watch your agent work
One prompt in your terminal. A finished image out. This loop is real output, generated through Cabina MCP.


Exactly how you use it
Same token, four ways to call Cabina. Pick a tab — copy, paste, run.
Create a key
Generate it in your profile and copy the token — shown once.
Drop in the token
Put it in an env var or a Bearer header — never hard-code it.
Call from anywhere
Terminal, your backend, a bot, or straight from Claude via MCP.
One key, a superpower per team
Same Cabina connection, a different job for every desk — from marketing to engineering.
Marketing & content ops
Auto-generate campaign copy, product descriptions and localized variants straight from your CMS or sheet.
Content automationCustomer support bots
Power Slack, Telegram or web-chat assistants that answer with your knowledge — backend-to-backend.
Custom integrationsAnalytics & reporting
Summarize datasets, draft weekly digests and explain numbers on a schedule, no analyst time required.
Automation scriptsHR & recruiting
Screen applications, draft outreach and generate job descriptions inside your ATS pipeline.
Internal toolsProduct & engineering
Embed Cabina in your app features, code-review bots and release notes — all from CI/CD pipelines.
CI/CD & backendSales & ops enablement
Generate proposals, follow-ups and CRM summaries automatically the moment a deal stage changes.
Workflow botsGood to know
All of them — 55+ models across text (GPT, Claude, Gemini, Llama, Grok, DeepSeek, Mistral), image (Midjourney, DALL·E, Flux, Stable Diffusion, Ideogram), video (Runway, Veo, Pika, Luma) and audio (ElevenLabs, Whisper). New models are added as they ship.
Yes. Once Cabina is connected, ask your agent in plain language — “use cabina to render a hero image with Flux” or “make a 10s clip with Veo” — and the asset comes back in your workflow.
No — MCP uses the same key and the same plan as the rest of Cabina. One subscription covers every model, instead of paying each vendor separately. There's a free tier to start.
A personal Bearer token that lets your code and external tools access Cabina without logging in through the browser. You can hold several keys and manage each independently.
The full token is shown only once — at the moment of creation. Copy and store it securely. If you lose it, simply regenerate the key.
No — it's the same key. Use any API key as the Bearer token in the MCP command to connect Cabina as an MCP server.
Yes. Rename, regenerate or delete any key at any time, and toggle keys between Active and Inactive without affecting the others.
Stop paying every AI vendor separately
A single Cabina key reaches GPT, Claude, Midjourney, Runway, ElevenLabs and 50+ more. Start on the free tier, scale only when you need to — and cancel anytime.
- 55+ models, one key
- Text, image, video & audio
- Works in any MCP client
Bring every model into your workflow
Create a key, copy the token, and start generating text, images and video from your agent — in under a minute.