Sunra MCP integrates image, video, audio and 3D generation into chat assistants

• Sunra brings image, video, audio and 3D generation into assistants including ChatGPT, Claude and Cursor, according to the launch announcement.
• Provider API prices with no markup and savings of up to 81% headline the launch thread. The effective cost depends on the model and credit plan.
• Seedance and Nano Banana are among the models promoted for Claude and Codex in AmirMushich's post.

Sunra's creative examples include a 150 BPM drift-phonk bed for a fencing ad and a rain-soaked taxi shot with follow-up camera directions. Its pricing comparison even names competitors that charge less for particular model settings.

Sunra uses MCP, the Model Context Protocol, to let an assistant call external generation tools. The launch video demonstrates image, audio and 3D generation inside a ChatGPT conversation.

The product page describes an agent choosing the model, writing the prompt and returning the result. Its examples span four categories:

• Images: GPT Image 2.5, Nano Banana Pro and Seedream 5.0 Pro, including workflows starting from uploaded photos.
• Video: Seedance 2.5, Kling 3.0 and Veo 3.1, with camera instructions and generated sound.
• Audio: ElevenLabs and MiniMax voices, music, sound effects and multilingual dubbing.
• 3D: Meshy 6, Meshy 7 and Tripo H3.1, producing textured GLB files from descriptions or photos.

These are Sunra's examples. AmirMushich, who promoted the integration for Claude and Codex, said he planned to start experimenting the following day.

Sunra's setup page gives chat assistants an OAuth connector and coding agents a CLI-based installation:

• Claude: Add Sunra through its connector link, then sign in and authorize access. The manual route is Customize → Connectors → Add custom connector.
• ChatGPT: Add a custom MCP server in Plugins using https://mcp.sunra.ai, select OAuth, then install the plugin from Personal and sign in.

For Claude Code, Codex and Cursor, Sunra publishes this three-step sequence:

1. Install: npm i -g @sunra/cli@0.2.2.
2. Authenticate: sunra login, followed by browser sign-in.
3. Inspect the account: sunra balance and sunra tools.

That launch-page flow requires no API key. The developer documentation separately describes a local @sunra/mcp-server package authenticated with SUNRA_KEY.

The official pricing explanation limits the maker-price promise to models whose providers publish an API price. Sunra converts that rate into credits at a nominal $0.01 each: a $0.40-per-second model consumes 40 credits per second.

The entry monthly subscriptions are:

• Basic: $12 for 1,200 credits.
• Pro: $29 for 3,500 credits.
• Creator: $125 for 17,000 credits.

A model consumes the same number of credits on every plan. Higher tiers and annual billing reduce the cash price of those credits; the same balance pays for generations in Sunra's web app and through the integration.

The advertised 81% is an “up to” claim. Sunra's annual comparison uses discounted Creator credits against competitors' annual-plan prices or lower recorded API prices, assumes all included credits are used, and excludes unlimited allowances.

The nicest workflow detail is the spending gate. Alongside MayorKingAI's promise of upfront pricing, the product FAQ says an agent can check the balance and estimate a job's cost before running it; credits are spent after confirmation unless auto-confirm is enabled.

Generated assets go into a project in the user's Sunra account. The assistant receives a link to each result, while the project keeps the outputs together.