pastehtml.dev MCP Server
Enables agents to publish and manage HTML/Markdown pastes, folders, and paste configuration via Model Context Protocol, with OAuth-based authentication and paste analytics.
README
<p align="center"> <a href="https://pastehtml.dev"> <img src="public/og-image.png" alt="pastehtml.dev — Share HTML in seconds" width="640"> </a> </p>
pastehtml.dev
Share HTML pages in seconds. Drop an HTML file, get a private link your friends can open, preview, and view the source of. Anonymous publishing still works, and accounts add folders, custom subdomains, view counts, browser-based re-upload updates, and account API keys for agent publishing into folders.
How it works
- Drag and drop (or browse for) an
.html/.htmfile up to 2 MB. - The document is published at
https://<token>.pastehtml.dev/— or at a claimed custom subdomain likehttps://launch-plan.pastehtml.dev/— with an inspector page (copy link, preview, highlighted source) at/p/<token>. Tokens are random 32-character lowercase-alphanumeric IDs (~165 bits of entropy — lowercase because they double as subdomains, and browsers lowercase hostnames), so random links are private unless shared. - Pastes can be password-protected. Locked pastes require the password before the inspector page, raw bytes, rendered preview, or live subdomain are served.
- Signed-in users can save pastes into folders, see view counts, and re-upload a
replacement
.html/.htmfile without changing the share link. Pastes can never be deleted. API-created update tokens still work for token-based automated updates. - The share page offers a live preview (with an in-site fullscreen mode) and a Rouge-highlighted source view.
- Every paste is served as a real page from its own origin —
https://<token>.pastehtml.dev/— so scripts and localStorage work (review-progress checklists persist) while staying fully isolated from other pastes and from the app itself. The path-based/p/<token>/rawendpoint returns the paste's exact bytes astext/plain— a byte-exact copy for programmatic clients, immune to any CDN HTML rewriting in transit — while/p/<token>/renderserves the same bytes astext/htmlinside a CSPsandbox.
Agent API
Publish without a browser — ideal for AI agents that produce design documents
or implementation plans and want to hand back a share link. Anonymous publishing
still works with no credentials; signed-in users can also create account API keys
at /api_keys so an agent can publish into their account and target folders.
# Multipart upload
curl -F "file=@plan.html" https://pastehtml.dev/api/pastes
# Or stream the HTML straight from a pipe
curl --data-binary @plan.html -H "Content-Type: text/html" \
"https://pastehtml.dev/api/pastes?filename=plan.html"
Anonymous creation returns 201 with
{ token, title, custom_subdomain, folder, owner, account_paste, password_protected, views_count, live_url, url, raw_url, render_url, markdown_url, update_token }
(or 422 with { errors }). Account-key creation returns the same paste fields
but omits update_token; update account-owned pastes with the account key so
revoking the key stops that agent's future access. Add password=<secret> (or
custom_subdomain=<name>, which requires an account key) to creation/update
requests; send clear_password=1 on updates to remove a paste password. Fetch raw_url to read a paste's exact bytes
back (text/plain, never rewritten in transit).
Account API keys for agents
A signed-in user can open API keys from the dashboard, create a key, and give
that secret to an agent. Use it as either Authorization: Bearer pht_... or
X-PasteHTML-API-Key: pht_.... When a valid account key is present on
POST /api/pastes, the paste is owned by that user and appears in their
dashboard. Folder targeting is account-only:
curl -H "Authorization: Bearer $PASTEHTML_API_KEY" \
--data-binary @plan.html -H "Content-Type: text/html" \
"https://pastehtml.dev/api/pastes?filename=plan.html&folder_name=Roadmap"
Use folder_name=<name> to find or create a folder under that account,
folder_id=<id> to target an existing folder, and clear_folder=1 on updates
to move a paste back to All pastes. Keys can also be scoped to a default folder
when created; scoped keys always publish there, can update only pastes already
in that folder, reject folder overrides, only list that folder through the folder
API, and cannot create unrelated folders. Scoped keys are revoked automatically
if the folder is deleted. Unscoped agents can discover existing folder IDs with:
curl -H "Authorization: Bearer $PASTEHTML_API_KEY" \
https://pastehtml.dev/api/folders
Unscoped keys can create folders with either nested form params
(folder[name]=Roadmap) or a simple top-level name=Roadmap, whichever is
easier for the client. Folder-scoped keys receive 403 on folder creation.
An account key can update pastes owned by that account, while the per-paste
update_token flow keeps working as long as the paste stays anonymous. If an
agent needs to claim an anonymous paste into an account, send the account key in
Authorization: Bearer ... and the paste secret separately as
X-Update-Token: .... Claiming is one-way: afterward the account key is the
paste's update credential and the old token stops working.
The update_token is revealed exactly once — the server stores only a digest —
and authorizes any number of in-place updates while the paste remains anonymous:
curl -X PATCH -H "Authorization: Bearer $UPDATE_TOKEN" \
-F "file=@plan.html" https://pastehtml.dev/api/pastes/$TOKEN
Updates accept the same two body forms as creation and return 200 with the
refreshed { token, title, custom_subdomain, folder, owner, account_paste, password_protected, views_count, live_url, url, raw_url, render_url, markdown_url },
403 for a wrong or missing update token or an account key that does not own the paste,
and 404 for an unknown paste. Paste publish/update endpoints are rate limited per IP at 20 requests per minute and 1,000 per day; folder API endpoints also require an account key and are rate limited per IP.
Agents discover all of this on their own: the full integration guide lives at
/llms.txt (also pointed to from the
homepage, both visibly and in an HTML comment for raw fetchers). Telling an
agent "publish this on pastehtml.dev" is enough.
MCP server
For agents that speak the Model Context Protocol,
pastehtml.dev is also a remote MCP server at https://pastehtml.dev/mcp
(Streamable HTTP). Instead of a pht_ key, the agent authorizes once through
your browser over OAuth and then works inside your account — the same folders,
view counts, and permanent pastes as the dashboard.
# Claude Code
claude mcp add --transport http pastehtml https://pastehtml.dev/mcp
# Codex
codex mcp add pastehtml --url https://pastehtml.dev/mcp
On first use the client opens a browser consent screen; approve it and the
agent is connected — no key to copy or store. Authorization is OAuth 2.1 with
PKCE and RFC 7591 Dynamic Client Registration, scoped to mcp:read,
mcp:pastes:write, and mcp:folders:write. Review or revoke connected
agents any time under Connected agents in the dashboard.
Ten tools are exposed (pastes are permanent — there is no delete-paste tool):
create_paste— publish a new HTML or Markdown paste, optionally into a folder.update_paste— republish an existing paste's content (overwrites it in place).configure_paste— change a paste's password, custom subdomain, or folder.get_paste— fetch one paste's metadata, URLs, and stored content.get_paste_stats— aggregate view analytics for a paste.list_pastes— page through the account's pastes, optionally filtered by folder.list_folders— list folders with their paste counts.create_folder— create a new, empty folder.rename_folder— rename a folder.delete_folder— delete a folder (its pastes survive, unfiled).
Dynamic Client Registration can be switched off in production with the
MCP_DYNAMIC_REGISTRATION_DISABLED environment variable (any already
pre-registered clients keep working). Smoke-test a deployment by fetching its
discovery document:
curl https://pastehtml.dev/.well-known/oauth-protected-resource
Stack
- Ruby on Rails 8.1 · PostgreSQL · Hotwire (Turbo + Stimulus)
- Tailwind CSS v4 (cssbundling) · esbuild (jsbundling) · Yarn 4
- Tooling via mise, PostgreSQL via Docker Compose
- Comic-book design: Bangers display type over Inter, halftone textures, ink-outlined panels with hard offset shadows
- Installable PWA: manifest + minimal network-first service worker with an offline fallback page (pastes themselves are never cached)
- SEO via meta-tags (OG/Twitter cards with a branded OG image, canonical, noindex on paste pages) and comic-styled static error pages
Development
mise run dev # starts postgres (docker), installs deps, prepares db, runs bin/dev
Or step by step:
mise run docker:start # postgres on localhost:5435
mise run deps # bundle install + yarn install
mise run db:prepare
bin/dev # rails server + js/css watchers on port 3000
Tests and lint
mise run test # rails test
mise run lint # rubocop + brakeman
Deployment
Deploys with Kamal from GitHub Actions (the
"Kamal Run" workflow) to a single server: the app container plus a postgres
18 accessory. db/production.sql creates the Solid Cache/Queue databases on
the accessory's first boot. Pastes are served from <token>.pastehtml.dev
subdomains, so the proxy routes the wildcard with a Cloudflare Origin CA
certificate (wildcards can't get Let's Encrypt certs over HTTP-01).
One-time setup:
- Cloudflare (free plan): add a proxied
Arecord forpastehtml.dev,www, and a proxied wildcard*record, all pointing at the server. Set SSL/TLS mode to Full (strict). Create an Origin Server certificate forpastehtml.dev, *.pastehtml.devand keep the PEM pair. - Repository Actions secrets:
SERVER_IP,SSH_PRIVATE_KEY(root access to the server),RAILS_MASTER_KEY(fromconfig/master.key),POSTGRES_PASSWORD,CLOUDFLARE_ORIGIN_CERTIFICATEandCLOUDFLARE_ORIGIN_KEY(the Origin CA PEM pair). The container registry (ghcr.io) authenticates with the workflow's ownGITHUB_TOKEN. - Run the Kamal Run workflow with the command
setuponce (provisions postgres + proxy), then withdeployfor every release.
For local runs of kamal commands, put the Origin CA pair in
.kamal/certs/origin.pem / origin-key.pem (gitignored), export
SERVER_IP and POSTGRES_PASSWORD, and remove the GHA builder cache block
from config/deploy.yml if you need to build the image locally.
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。
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。
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。
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。
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。
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。
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。