C
contentworker/contentworker
API-first, AI-agent-native headless CMS, edge-native on Cloudflare: Workers, D1, Durable Objects, Workflows,…
在 GitHub 查看 ↗链接已于 2026-11-12 核查 · HTTP 200 · 打开的是厂商自己的官网
安装与配置
客户端配置（取自该仓库的安装命令）
{
"mcpServers": {
"contentworker-contentworker": {
"command": "npx",
"args": ["wrangler"]
}
}
}解析自我们于 2026-11-01 抓取的该仓库 README；我们未执行过。
command/args 来自仓库写明的安装命令；不确定时请以该项目 README 为准。
VS Code（.vscode/mcp.json —— 键名是 servers）
{
"servers": {
"contentworker-contentworker": {
"command": "npx",
"args": ["wrangler"]
}
}
}VS Code 顶层键是 <code>servers</code>，其余与上面那份相同。
Claude Code（命令行）
claude mcp add --transport stdio contentworker-contentworker -- npx wrangler
Claude Code 不收 JSON 文件，用这条命令直接加。
配置项的键名以你的客户端为准：客户端与 CLI
- ★ 数取值
- 0 ★ · contentworker/contentworker · 取自 GitHub 搜索 API · （2026-11-13 20:00:50 核对）
- 来源明细
- 从该仓库的 GitHub 页面直接读取 · 条目类型：开源项目
- 收录与活跃
- 收录于 2026-11-02 · 最近见到 2026-11-14 · 共见到 13 次 · 被 3 个来源列出
- 评分构成
- 综合分 303 · 3 个来源 · repo
- 安装
npx wrangler
常见问题（答案都来自我们的记录）
怎么安装？
该仓库给出的安装命令是
npx wrangler。我们的记录里它是仓库自己提供的；不保证在你的客户端上直接可用。它适合谁 / 用来做什么？
我们把它归在这些标签下: 开源项目。标签来自条目自身，不是我们强加的分类。
谁维护它、还在更新吗？
仓库
contentworker/contentworker；我们收录于 2026-11-02；最近见到 2026-11-14。这里没有"官方认证"或"质量评级"——我们只记录来源与时间。# contentworker
以下摘录自该仓库的 README（片段） · 在 GitHub 读完整 README →
A composable, API-first, AI-agentic-first headless CMS: the AI surface is a first-class
client rather than a bolt-on. Model structured content once and deliver it anywhere (web, mobile,
kiosks, email).
It runs entirely on Cloudflare. D1 is the system of record, with Workers, Queues, KV, R2,
Durable Objects, Workflows, Vectorize, Workers AI (through AI Gateway), Images, Stream, Email
Sending and Worker Loader behind it. The core is strict ports & adapters, so every platform
service sits behind an interface and the business logic never touches a binding. Each customer
gets one isolated deployment (its own D1, R2, KV, queues and Workers); spaces and environments
live inside that deployment's D1.
## Highlights
Workers split by traffic profile. The admin (apps/web), the Management and Preview API
(apps/api), the public Delivery API (apps/delivery) and the asynchronous engine
(apps/engine) are separate Workers, each with its own CPU limit, placement and failure blast
radius.
Hexagonal core. domain depends on nothing; application depends only on ports;
Cloudflare adapters (@contentworker/adapters, one subpath each) and the D1 store
(@contentworker/db) are bound in one place, @contentworker/platform's wire.ts.
One set of use-cases, every caller. The HTTP API, the MCP server (165 tools), the queue
consumers and the agents call the same application use-cases, so an AI agent can never do
what a human API client couldn't, and every write goes through the same validation and RBAC.
Structured content model. Typed fields with validations, localization with fallback
chains, references with referential integrity, components, and assets in R2.
Three API surfaces. Management (CMA), Delivery (CDA) and Preview (CPA), each gated by
scoped, hashed API keys and roles with per-type and per-field grants. Delivery is REST plus a
per-environment GraphQL schema generated from the model, served from a cache keyed by what
the caller may see; the key is checked on every request, before the cache.
Event-driven backbone. Every write commits one D1 batch that includes its domain event in a
transactional outbox. A post-commit nudge (plus a one-minute cron sweep) relays it to Queues,
whose engine consumer fans it out to webhooks, functions, search indexing, live SSE, delivery
cache purges and the agents.
AI built in. Tiered model selection (flagship/balanced/fast) on Workers AI,
schema-validated generation, hybrid search (FTS5 shards in Durable Objects plus Vectorize),
and durable enrich/moderate/curate/repurpose agents and editorial workflows on Cloudflare
Workflows.
Sandboxed plugins. Third-party write-path plugins run in per-plugin Worker Loader isolates
with no network and only the capabilities a person approved.
## Documentation
The public docs (guides for editors and integrators) are the Starlight site in apps/docs. The
maintainer docs live in docs/ (start at the index):
| Doc | What's in it |
| Architecture | The Workers, the hexagonal layers, the package map, request and event flows |
| Domain model | Field types, validation, entry state machine, locales, references, events |
| API reference | Every HTTP endpoint, the query language, hybrid search, live SSE |
client rather than a bolt-on. Model structured content once and deliver it anywhere (web, mobile,
kiosks, email).
It runs entirely on Cloudflare. D1 is the system of record, with Workers, Queues, KV, R2,
Durable Objects, Workflows, Vectorize, Workers AI (through AI Gateway), Images, Stream, Email
Sending and Worker Loader behind it. The core is strict ports & adapters, so every platform
service sits behind an interface and the business logic never touches a binding. Each customer
gets one isolated deployment (its own D1, R2, KV, queues and Workers); spaces and environments
live inside that deployment's D1.
## Highlights
Workers split by traffic profile. The admin (apps/web), the Management and Preview API
(apps/api), the public Delivery API (apps/delivery) and the asynchronous engine
(apps/engine) are separate Workers, each with its own CPU limit, placement and failure blast
radius.
Hexagonal core. domain depends on nothing; application depends only on ports;
Cloudflare adapters (@contentworker/adapters, one subpath each) and the D1 store
(@contentworker/db) are bound in one place, @contentworker/platform's wire.ts.
One set of use-cases, every caller. The HTTP API, the MCP server (165 tools), the queue
consumers and the agents call the same application use-cases, so an AI agent can never do
what a human API client couldn't, and every write goes through the same validation and RBAC.
Structured content model. Typed fields with validations, localization with fallback
chains, references with referential integrity, components, and assets in R2.
Three API surfaces. Management (CMA), Delivery (CDA) and Preview (CPA), each gated by
scoped, hashed API keys and roles with per-type and per-field grants. Delivery is REST plus a
per-environment GraphQL schema generated from the model, served from a cache keyed by what
the caller may see; the key is checked on every request, before the cache.
Event-driven backbone. Every write commits one D1 batch that includes its domain event in a
transactional outbox. A post-commit nudge (plus a one-minute cron sweep) relays it to Queues,
whose engine consumer fans it out to webhooks, functions, search indexing, live SSE, delivery
cache purges and the agents.
AI built in. Tiered model selection (flagship/balanced/fast) on Workers AI,
schema-validated generation, hybrid search (FTS5 shards in Durable Objects plus Vectorize),
and durable enrich/moderate/curate/repurpose agents and editorial workflows on Cloudflare
Workflows.
Sandboxed plugins. Third-party write-path plugins run in per-plugin Worker Loader isolates
with no network and only the capabilities a person approved.
## Documentation
The public docs (guides for editors and integrators) are the Starlight site in apps/docs. The
maintainer docs live in docs/ (start at the index):
| Doc | What's in it |
| Architecture | The Workers, the hexagonal layers, the package map, request and event flows |
| Domain model | Field types, validation, entry state machine, locales, references, events |
| API reference | Every HTTP endpoint, the query language, hybrid search, live SSE |
收录徽章（可选）
想把收录标记放进 README 就复制这行，可选，不影响收录。
[](https://anclum.com/zh/s/contentworker-contentworker/)
- 来源
- github · https://github.com/contentworker/contentworker
- 按领域浏览
- 开源项目
- 收录来源
- 3 个独立来源
- 收录于
- 今天
同类推荐
所属领域与标签
每个标签都能点进去看同类的全部条目。数字是当前库里的真实条数。