starlight-creator-mcp
Enables agentic creators to generate images, videos, and audio, run taste-based scoring, and publish through human-gated signed manifests using their own keys and local provenance.
README
starlight-creator-mcp
One MCP engine for agentic creators, built so the user's own machine and keys do all the work. Brand packs (Arcanea, GenCreator, Author OS, Anime Legends) sit on top as data; this repository is the unbranded substrate they share.
- BYOK, client-side. Provider keys live in your OS keychain. Nothing here proxies, logs or stores them.
- Local ledger with provenance. Every generated or imported file is content-addressed (sha256) and carries a
.provenance.jsonsidecar: model, prompt, seed, cost, pack, rubric scores. - Taste loop.
best_of/min_scoregenerate N candidates, score each against the loaded pack's rubric with a vision judge, keep the best. - Human-gated publishing. Posts go through a signed manifest: prepare → the person types an 8-character confirm code → HMAC-signed → publish through your Postiz or a local outbox. Agents cannot self-approve. Numbers without evidence are blocked.
- Packs are data. Canon, rubrics, templates, prompts, styles and routing preferences come from
pack.jsondirectories; two packs compose (e.g.author-os+arcanea).
Install
The npm name starlight-creator-mcp is not on the registry yet (verified 2026-08-22: packument 404). Do not run npx -y starlight-creator-mcp until a first publish exists. Use this repo:
git clone https://github.com/frankxai/starlight-creator-mcp.git
cd starlight-creator-mcp
pnpm install
pnpm -r --filter './packages/**' run build
node packages/bundle/dist/main.js doctor
node packages/bundle/dist/main.js setup --print # prints Claude / Cursor mcpServers JSON
setup without --print is interactive (stores keys in the OS keychain). Non-TTY correctly exits 2.
Verified locally 2026-08-22 (main 4185596): doctor exit 0 with starter pack + local ledger/storage; keys optional.
After the first publish, the install path becomes npx -y starlight-creator-mcp. Until then, point MCP clients at node /absolute/path/to/packages/bundle/dist/main.js.
Providers in this tree: OpenRouter, MuAPI, Google Gemini. fal and ElevenLabs are optional later adapters, not required to run doctor.
Tools (16)
creator_models · creator_generate_image · creator_generate_video · creator_generate_audio · creator_transcribe · creator_job_status · creator_score · creator_assets · creator_asset_sync · creator_pack_info · creator_publication_prepare · creator_publication_approve · creator_publish · creator_metrics_ingest · creator_metrics_query · creator_license_status
Every tool has an input schema, an output schema (structuredContent) and annotations. Long jobs return a job_id within wait_ms (≤ 20 s) instead of blocking. Resources: pack://<id>/manifest, pack://<id>/canon/<file>, pack://<id>/templates/<id>, pack://<id>/rubrics/<id>, creator://assets/<sha>, creator://jobs/<id>. Pack prompts become MCP prompts.
Packages
| Package | Role |
|---|---|
@starlight-intelligence/creator-core |
providers, router, taste loop, vault, fetch guard, sandbox, storage, ledger, publishing, license, packs |
@starlight-intelligence/creator-server |
MCP server factory (createCreatorServer) + tools/resources/prompts |
@starlight-intelligence/creator-packs-spec |
creator-pack-manifest.v1 schema, validator, merge rules, creator-pack CLI |
@starlight-intelligence/creator-cli |
`creator-mcp stdio |
@starlight-intelligence/creator-provider-{openrouter,muapi,google} |
provider adapters |
starlight-creator-mcp |
the installable bundle |
Build a brand on it
import { createCreatorServer } from '@starlight-intelligence/creator-server'
import { createOpenRouterProvider } from '@starlight-intelligence/creator-provider-openrouter'
export const createArcaneaServer = () => createCreatorServer({
name: 'arcanea-mcp', version: '1.0.0',
packs: [PACK_DIR], // data only: canon, rubrics, prompts, styles
providers: vault => [createOpenRouterProvider({ vault })],
tools: { alias: { creator_generate_image: 'arcanea_generate_image' }, defaults: { creator_generate_image: { style: 'arcanean-cinematic', best_of: 2 } } },
})
Rules: brands depend on the substrate, never on each other; packs are data, not code; one direction of dependency.
Security model
Read SECURITY.md. Short version: keys in the keychain, every outbound fetch goes through a host allowlist that blocks private ranges, every write is sandboxed, publishing requires a human-entered code, and the server never executes shell strings.
Develop
pnpm install
pnpm -r --filter './packages/**' run build
pnpm test # unit + protocol-level e2e (spawns the stdio server, mocks providers)
pnpm lint
pnpm inspect # MCP Inspector against dist/stdio.js
MIT.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。