Summary
Gives Claude access to 14 image and video generation models through a single OAuth-secured endpoint. You get text-to-image, image editing, text-to-video, and image-to-video with models ranging from cheap options like z_image at 10 coins to premium ones like SeeDance 2 at up to 12,240 coins. Four tools let you list models, start generation jobs, and poll for results. The async pattern returns a job_id immediately, then you call check_job until it completes. Small images come back inline as base64, larger assets and videos return Firebase Storage URLs. Same backend runs the Dora iOS app and web product. Reach for this when you want Claude to generate visual content without managing individual API keys for each model provider.
Keeps your Mac awake while Claude Code, Codex or Cursor works. Lets it sleep when they are done.
Try free for 7 days →
Block distracting apps from your iPhone permanently without a 3rd party app. Free and open source.
Block now (100% free) →
Notes, actions and memory. Without a meeting bot. First month 100% off.
Download for free →
belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →
Free an open-source skills that make AI coding agents easier to understand, verify, and control.
Download for free →
Connect your Claude agent to live crypto prices and trading routes via 1inch
Get the MCP →Tools
Public tool metadata for what this MCP can expose to an agent.
4 tools
list_modelsList Dora generation models, cheapest-first, with per-call coin cost. Call this when the user asks what models are available, when cost matters, or when you want to verify a model name. Defaults: nano_banana_2 for images, seedance_1_5_pro for videos — only step up if the user...
List Dora generation models, cheapest-first, with per-call coin cost. Call this when the user asks what models are available, when cost matters, or when you want to verify a model name. Defaults: nano_banana_2 for images, seedance_1_5_pro for videos — only step up if the user...
No parameter schema in public metadata yet.
generate_videoGenerate a video. Default model: seedance_1_5_pro at 480p, 8s, no audio (cheapest reasonable quality). Before calling, if the user has not specified them, briefly ASK for: (1) aspect ratio (9:16 vertical, 16:9 horizontal, 1:1 square), (2) duration (4/8/12s), (3) whether they w...7 params
Generate a video. Default model: seedance_1_5_pro at 480p, 8s, no audio (cheapest reasonable quality). Before calling, if the user has not specified them, briefly ASK for: (1) aspect ratio (9:16 vertical, 16:9 horizontal, 1:1 square), (2) duration (4/8/12s), (3) whether they w...
Parameters* required
aspect_ratiostringone of
16:9 · 9:16 · 1:1durationnumberone of
4 · 5 · 6 · 8 · 10 · 12image_urlstringmodelstringone of
seedance_1_5_pro · grok_text_to_video · grok_image_to_video · kling_2_6 · veo_3 · seedance_2_fastpromptstringresolutionstring480p cheapest; 1080p ~5x cost. Default 480p.one of
480p · 720p · 1080pwith_audiobooleanGenerate sound. SeeDance/Kling support audio — typically doubles cost. Default false.
generate_imageGenerate an image. Default model: nano_banana_2 (good quality, ~100 coins, supports reference photo). Use z_image (10 coins) only if the user wants the cheapest possible result. Use nano_banana_pro (180-240) only when the user explicitly asks for face preservation or maximum q...5 params
Generate an image. Default model: nano_banana_2 (good quality, ~100 coins, supports reference photo). Use z_image (10 coins) only if the user wants the cheapest possible result. Use nano_banana_pro (180-240) only when the user explicitly asks for face preservation or maximum q...
Parameters* required
aspect_ratiostringone of
16:9 · 9:16 · 1:1modelstringone of
nano_banana_2 · z_image · google_nano_banana · google_nano_banana_edit · gpt4o_image · gpt_image_2promptstringreference_image_urlstringresolutionstring1K cheapest; 4K up to 50% more. Default 1K.one of
1K · 2K · 4Kcheck_jobCheck the status of a generation job. Returns status (pending|done|failed) and result_url when done. Wait at least 30 seconds between polls — generations take 1-5 minutes typically.1 params
Check the status of a generation job. Returns status (pending|done|failed) and result_url when done. Wait at least 30 seconds between polls — generations take 1-5 minutes typically.
Parameters* required
job_idstringDora Video MCP — AI Video Generation MCP Server
Dora Video is a hosted AI video and image generation MCP server for ChatGPT, Claude, Cursor, Google Antigravity and any other Model Context Protocol client. One OAuth-secured connection gives an agent 16 models — Google Veo 3, Kling 2.6, SeeDance 2, SeeDance 1.5 Pro, Grok Imagine, Nano Banana Pro and others — with the coin cost of each returned before anything is generated, sorted cheapest first.
https://doravideo.com/mcp
Built and operated by Saga Labs. Full public specification, limitations included: https://doravideo.com/ai-video-mcp
Why this exists
Most video MCP servers are either a single provider, or raw infrastructure that hands your agent a job id and leaves it to build polling, retries and billing. Dora sits in between: a curated catalogue behind one account and one balance, four tools an agent can reason about in a single pass, and a finished MP4 or image at the end.
The design bet is that an agent picks better from a short list. A tool per model, or per provider, makes the tool list longer than the task.
Tools
| Tool | What it does |
|---|---|
list_models | Every model with its live coin cost, max duration, and whether it accepts an image or a video input. Sorted cheapest first. |
generate_video | Text-to-video, image-to-video, or motion transfer from a driving clip. Returns a job id. |
generate_image | Text-to-image and image editing from reference photos. Returns a job id. |
check_job | Polls a job and returns the finished file. |
Because list_models publishes prices, this works:
"Animate this product photo as a vertical ad. Use the cheapest model that can do it, tell me what it will cost, and wait for me to confirm."
Models
Video (9): grok_1_5, seedance_1_5_pro, grok_image_to_video,
grok_text_to_video, kling_2_6, kling_2_6_motion, veo_3,
seedance_2_fast, seedance_2
Image (7): z_image, google_nano_banana, google_nano_banana_edit,
gpt4o_image, gpt_image_2, nano_banana_2, nano_banana_pro
Costs
1 Dora coin = 1 provider credit, no multiplier. Dollar figures are the coin value on the Pro plan ($15 for 1,000 coins). Verified 5 September 2026.
| Model | Configuration | Coins | On Pro |
|---|---|---|---|
z_image | 1 image | 1 | $0.02 |
google_nano_banana | 1 image | 4 | $0.06 |
nano_banana_2 | 1K–2K image | 8 | $0.12 |
grok_1_5 | 4s · 480p video, native audio | 10 | $0.15 |
seedance_1_5_pro | 4s · 480p video | 14 | $0.21 |
nano_banana_pro | 1K–2K image | 18 | $0.27 |
kling_2_6 | 5s video, no audio | 55 | $0.83 |
veo_3 | 8s video (Veo 3 Fast) | 60 | $0.90 |
seedance_2 | 4s · 480p video | 76 | $1.14 |
Longer durations, higher resolutions and audio scale up. list_models is
always authoritative; this table is a snapshot.
Connect it
Remote (ChatGPT, Claude, Cursor, Antigravity)
Add https://doravideo.com/mcp as a remote MCP server and approve the OAuth
prompt. OAuth 2.1 with dynamic client registration; no API key to manage.
Step-by-step guides:
Stdio bridge
For clients that only speak stdio, bin/dora-mcp.js in this repo bridges to the
same endpoint with an API key from
doravideo.com/account:
{
"mcpServers": {
"dora": {
"command": "node",
"args": ["/path/to/dora-mcp/bin/dora-mcp.js"],
"env": { "DORA_API_KEY": "dora_sk_..." }
}
}
}
What Dora does not do
Stated here so nobody discovers it later:
- No public REST API today. Access is the web app, the iOS app, or MCP.
- No deterministic price router.
list_modelsreturns costs cheapest first and the agent chooses; Dora does not choose for it. list_modelsreturns the default configuration cost, not a quote for every parameter combination.- No publishing or scheduling. Dora returns files; you post them.
- Generating requires a paid plan or a credit pack. There is no free tier.
If you are building your own high-volume video backend, model infrastructure platforms are a better fit. Dora is a better fit when the goal is a finished clip with minimal integration.
Pricing
| Plan | Price | Credits |
|---|---|---|
| Pro (web) | $15 / month | 1,000 / month |
| Credit packs | from $7 | 600+, never expire |
Every plan includes every model, every resolution, commercial rights and no watermark.
Also available without an agent
The same account and balance work on the web at doravideo.com and on the iPhone app. Android is in review.
Links
- Specification: https://doravideo.com/ai-video-mcp
- Model catalogue: https://doravideo.com/models
- Studio: https://aisagalab.com
- Support: https://doravideo.com/support · sagalabs@proton.me
The MCP server itself is hosted and closed-source. This repository is the public documentation and the stdio bridge.
Featured
Keeps your Mac awake while Claude Code, Codex or Cursor works. Lets it sleep when they are done.
Try free for 7 days →Block distracting apps from your iPhone permanently without a 3rd party app. Free and open source.
Block now (100% free) →Notes, actions and memory. Without a meeting bot. First month 100% off.
Download for free →belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →Free an open-source skills that make AI coding agents easier to understand, verify, and control.
Download for free →Connect your Claude agent to live crypto prices and trading routes via 1inch
Get the MCP →
