opensora2-prompt-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@opensora2-prompt-mcp Build a 3-shot video prompt for a red balloon drifting over Paris."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
OpenSora 2 Prompt MCP
A deterministic, read-only MCP server for designing video prompts and shot plans. It complements the independent resources at OpenSora2.com and does not generate videos, call model providers, access accounts, or spend credits.
Tools
build_video_promptturns a rough idea into a structured prompt pack.plan_shotscreates a timed 1–6 shot sequence with continuity anchors.diagnose_promptfinds missing motion, camera, lighting, and continuity controls.get_resourcesreturns canonical OpenSora2.com guides and workflow pages.
All tools are deterministic and declare read-only MCP annotations. No API key is required.
Related MCP server: seevido
Install
Run directly from GitHub:
{
"mcpServers": {
"opensora2": {
"command": "npx",
"args": ["-y", "github:gpt-img-2/opensora2-prompt-mcp"]
}
}
}Or clone and run locally:
pnpm install
pnpm build
node dist/index.jsOptional environment variable:
OPENSORA2_APP_BASE_URL: changes the resource-link origin. Defaults tohttps://opensora2.com.
Example inputs
Build a prompt:
{
"idea": "A red ceramic teapot steams on a rain-streaked cafe window",
"workflow": "image-to-video",
"camera": "slow 10 cm push-in",
"motion": "steam curls upward while raindrops slide down the glass",
"referenceConstraints": "teapot silhouette, glaze pattern, and window composition"
}Plan shots:
{
"idea": "A trail shoe crosses wet rock and lands in a shallow stream",
"shotCount": 3,
"totalDurationSeconds": 9,
"continuityAnchor": "shoe color, laces, runner wardrobe, and travel direction"
}OpenClaw Skill
The companion Skill is in openclaw/opensora-video-prompt-architect. It works as a text-only workflow without this MCP; connecting the MCP adds deterministic prompt-building and diagnostic tools.
Development
pnpm validate
pnpm test:smokeThis project is an independent prompt-design utility and is not the official Open-Sora project or an official model implementation.
License
MIT
Available Tools
4 toolsbuild_video_promptBuild an OpenSora video promptBRead-onlyIdempotent
Turn a rough video idea into a deterministic prompt pack with motion, camera, lighting, continuity constraints, and revision moves. This does not generate video.
Inputs: idea (required), style, camera, motion, +6 more.
diagnose_promptDiagnose a video promptARead-onlyIdempotent
Check whether a video prompt contains a readable scene, visible motion, camera direction, lighting, and continuity controls. Returns focused revision suggestions.
Inputs: prompt (required).
get_resourcesGet OpenSora 2 resourcesARead-onlyIdempotent
Return canonical OpenSora2.com prompt, workflow, technical, or safety resources. Does not fetch external content.
Inputs: topic.
plan_shotsPlan a video shot sequenceARead-onlyIdempotent
Break one video idea into 1 to 6 timed shots with purpose, camera direction, and continuity anchors. Read-only and deterministic.
Inputs: idea (required), shotCount, cameraStyle, continuityAnchor, +1 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 4 tools
Each tool addresses a distinct phase of the video prompt workflow: fetching reference resources, constructing a prompt, breaking ideas into shots, and diagnosing prompt quality. There is no overlap in purpose, making selection unambiguous.
All tool names follow a consistent verb_noun pattern: get_resources, build_video_prompt, plan_shots, diagnose_prompt. This uniform structure makes it easy to predict tool functions.
Four tools is a well-scoped set for a focused video prompt crafting server. Each tool is essential to the core workflow and avoids unnecessary bloat.
The server covers the main lifecycle of prompt creation: resource lookup, prompt building, shot planning, and diagnosis. Minor gaps exist (e.g., no explicit revision application tool), but agents can apply suggested changes manually, so the surface is largely complete.
Maintenance
Related MCP Connectors
Free, open source library of AI video prompts that really worked, each next to the video it produced and credited to its creator. Describe a video in plain words, in any language: your AI finds the prompts that fit by meaning, shows you the videos and can adapt the best one to your idea. Read only, with no account and no key.
Design, save, and run outcome-aligned AI workflows and verifiers, with reliable image output.
Plan AI video shots: model picks, sourced prompts, prompting guides and real $/s costs.
Create and edit AI videos from chat: plan shots, generate scenes, and export stories and ads.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables deterministic, read-only creation of bilingual video prompts, timed reference-aware shot plans, and focused prompt diagnostics without API keys or external provider calls.4MIT
- AlicenseAqualityDmaintenanceProvides read-only tools to build English or Chinese video prompts, plan reference-aware sequences, diagnose prompts, and fetch canonical SeeVido resources without API keys.4MIT
- AlicenseAqualityDmaintenanceEnables deterministic, read-only construction of Seedance-oriented video prompts, planning of multimodal reference-aware sequences, prompt diagnosis, and discovery of canonical See Dance 2 resources.4MIT
- AlicenseAqualityCmaintenanceA read-only, deterministic server that enables assembling copy-ready video prompts, planning timed shot sequences, diagnosing prompt clarity and current duration/resolution/reference/end-frame/audio limits, and retrieving canonical creation, guide, pricing, and policy URLs.4MIT