io.github.CakeRepository/1password
编码与调试by cakerepository
提供用于 1Password service accounts 的 MCP server,支持安全集成与凭据访问管理。
什么是 io.github.CakeRepository/1password?
提供用于 1Password service accounts 的 MCP server,支持安全集成与凭据访问管理。
README
1Password MCP Server
Give your AI assistant a locked door to 1Password — not a pile of passwords in the chat.
@takescake/1password-mcp is a community Model Context Protocol (MCP) server. It lets Claude, Cursor, VS Code Copilot, OpenAI Codex, Gemini, and other MCP clients manage vaults and credentials through a 1Password Service Account.
Built on the MCP TypeScript SDK v2 with protocol negotiation for 2026-07-28 (and legacy clients). Secrets stay in 1Password; agents prefer op:// references and op_run so plaintext never has to land in the model transcript.
Not an official 1Password product. Community-built, Apache 2.0 licensed.
Who this is for
| You are… | You get… |
|---|---|
| Not deeply technical | A one-time setup: create a service account, paste the token into your AI app’s MCP settings, then ask in plain English (“list my automation vault”, “rotate the staging DB password”). |
| An agent / LLM | Fifteen typed tools, four workflow prompts, three browsable resources, and clear rules: prefer op_run + op:// over revealing secrets. |
| A developer / SRE | Full item CRUD, secure notes, password generation, vault allow-lists, Keychain token loading on macOS, and CI-friendly env config. |
Why teams pick this server
- Security-first defaults —
password_readanditem_getreturn metadata unless you opt in withreveal: true. op_run(the MCP equivalent ofop run) — injectop://vault/item/fieldinto a local command’s environment; plaintext is redacted from stdout/stderr and never logged back to the model.- Full vault toolkit — list, search, get, edit, create logins & notes, rotate passwords, archive, or delete.
- Guided prompts — password generation, credential rotation, vault audit, and secret-reference helpers.
- Browsable resources — vault and item catalogs over
1password://…URIs (no secrets in resource payloads). - Modern MCP — stdio transport, Zod 4 schemas, MCP 2026-07-28 negotiation with legacy client compatibility.
What you get
Tools (15)
Grouped the way agents and humans actually use them.
Discover
| Tool | What it does |
|---|---|
vault_list | List vaults the service account can access (id, name, description, type). |
item_lookup | Search a vault by title substring; optional limit (max 200). |
item_list | List every item in a vault (id, title, category, tags, updatedAt) — never secrets. |
Read (safe by default)
| Tool | What it does |
|---|---|
item_get | Full item: title, category, tags, notes, fields. Concealed values stay hidden unless reveal: true. Accepts op://… or vaultId + itemId. |
password_read | Read one field (default password) via op://… or ids. Metadata-only unless reveal: true. Prefer op_run to use a secret. |
op_check_ref | Validate op://vault/item/field and return non-secret metadata only (vault, item, field). Never the value. |
Create & update
| Tool | What it does |
|---|---|
password_create | Create a Login or Password item (username, URL, tags, notes). returnSecret defaults to false. |
note_create | Create a Secure Note with optional tags and custom fields. |
password_update | Rotate a password / concealed field (creates the field if missing). |
item_edit | Update title, notes (empty string clears), tags, URL; upsert or remove fields. Unreferenced fields stay untouched. |
password_generate | Cryptographically secure random password (length 8–128; symbols/numbers/uppercase toggles). |
password_generate_memorable | Memorable passphrase from a ~500-word list (word count, separator, number/symbol suffixes). |
Use secrets without revealing them
| Tool | What it does |
|---|---|
op_run | Run a local command (command or argv) with env vars. Values matching op://… are resolved into the child process only; resolved secrets are redacted from returned output. Optional cwd, shell, timeout_ms, stdin. |
Soft-delete & destroy
| Tool | What it does |
|---|---|
item_archive | Move an item to the archive (hidden from normal views). |
item_delete | Permanently delete an item — cannot be undone. |
Prompts (4)
| Prompt | When to use it |
|---|---|
generate-secure-password | Generate (random or memorable) and optionally store — without dumping the password into chat. |
credential-rotation | Find → verify access → generate → update → confirm op:// reference. |
vault-audit | Inventory a vault by category; flag duplicates / oddities — never reveal secrets. |
secret-reference-helper | Build a paste-ready op://vault/item/field from names. |
Resources (3)
| URI | Contents |
|---|---|
1password://config | Non-secret server config (name, version, log level, token source, Node version). |
1password://vaults | JSON list of accessible vaults. |
1password://vaults/{vaultId}/items | JSON item metadata for one vault (no secret values). |
Before you start
You need two things:
- Node.js 20 or newer
- A 1Password Service Account with access to the vault(s) you want the AI to use
Create a service account (plain English)
- Sign in to your 1Password account on the web.
- Open Developer → Service Accounts (or follow 1Password’s guide).
- Create a service account and grant it only the vaults you want automation to touch (for example an
AutomationorCIvault — not your personal banking vault). - Copy the token once. Treat it like a master key.
Quick start
Claude Desktop / Cursor / VS Code / most IDEs
Add this to your MCP config (exact file depends on the app):
{
"mcpServers": {
"1password": {
"command": "npx",
"args": ["-y", "@takescake/1password-mcp"],
"env": {
"OP_SERVICE_ACCOUNT_TOKEN": "YOUR_SERVICE_ACCOUNT_TOKEN"
}
}
}
}
Restart the app, then try: “List my 1Password vaults.”
macOS Keychain (no token in the config file)
Store the token in Keychain, then point the server at it:
{
"mcpServers": {
"1password": {
"command": "npx",
"args": ["-y", "@takescake/1password-mcp"],
"env": {
"OP_KEYCHAIN_SERVICE": "op-service-account-claude-automation",
"OP_KEYCHAIN_ACCOUNT": "your-macos-username"
}
}
}
}
Token resolution order: CLI (--service-account-token / --token) → OP_SERVICE_ACCOUNT_TOKEN → macOS Keychain. OP_KEYCHAIN_ACCOUNT is optional when the service name alone is unique.
OpenAI Codex (TOML)
Option A — token in config:
[mcp_servers."1password"]
command = "npx"
args = ["-y", "@takescake/1password-mcp"]
[mcp_servers."1password".env]
OP_SERVICE_ACCOUNT_TOKEN = "YOUR_SERVICE_ACCOUNT_TOKEN"
Option B (recommended) — config only names the env var:
[mcp_servers."1password"]
command = "npx"
args = ["-y", "@takescake/1password-mcp"]
env_vars = ["OP_SERVICE_ACCOUNT_TOKEN"]
Set OP_SERVICE_ACCOUNT_TOKEN in your shell or CI. Note: codex mcp add ... --env OP_SERVICE_ACCOUNT_TOKEN=... writes the secret into Codex config; prefer env_vars when you can.
On macOS you can omit the token env and use OP_KEYCHAIN_SERVICE (+ optional OP_KEYCHAIN_ACCOUNT) instead.
Optional: lock op_run / op_check_ref to certain vaults
By default those tools may resolve op:// references from any vault the service account can see. To allow-list vaults:
{
"env": {
"OP_SERVICE_ACCOUNT_TOKEN": "YOUR_SERVICE_ACCOUNT_TOKEN",
"OP_MCP_ALLOWED_VAULTS": "Automation, CI"
}
}
Names or IDs work. References outside the list are rejected before resolution. Same setting via --allowed-vaults.
For agents: how to handle secrets
Follow this order every time:
- Discover with
vault_list→item_lookup/item_list(metadata only). - Confirm a reference with
op_check_ref— neverrevealjust to see if a path exists. - Use a secret in a command or API call with
op_runandop://vault/item/fieldinenv. - Reveal with
password_read/item_get+reveal: trueonly when the human explicitly needs the value in chat. - Rotate with
password_generate→password_update(keepreturnSecret: falseunless asked). - Prefer
item_archiveoveritem_deleteunless permanent removal is required.
op_run sketch
{
"argv": ["curl", "-sS", "https://api.example.com/health"],
"env": {
"API_TOKEN": "op://Automation/Example API/credential"
},
"timeout_ms": 60000
}
Prefer argv over a shell command string when you can — fewer quoting surprises.
Configuration reference
Environment variables
| Variable | Required | Description |
|---|---|---|
OP_SERVICE_ACCOUNT_TOKEN | Usually yes | Service account token. Not required on macOS if Keychain vars are set. |
OP_KEYCHAIN_SERVICE | No | macOS: Keychain service name for the token. |
OP_KEYCHAIN_ACCOUNT | No | macOS: optional account to narrow the Keychain lookup. |
OP_MCP_ALLOWED_VAULTS | No | Comma-separated vault names/IDs allowed for op_run / op_check_ref. Empty = unrestricted. |
OP_INTEGRATION_NAME | No | Name reported to the 1Password SDK (default: 1password-mcp). |
OP_INTEGRATION_VERSION | No | Version reported to the SDK (default: package version). |
MCP_LOG_LEVEL | No | debug | info | warn | error (default: info). |
MCP_DEBUG | No | If set, forces debug logging. |
CLI flags
--service-account-token <token> 1Password service account token
--token <token> Alias for --service-account-token
--log-level <level> error | warn | info | debug (default: info)
--integration-name <name> Custom integration name for the 1Password SDK
--integration-version <version> Custom integration version
--allowed-vaults <list> Comma-separated allow-list for op_run / op_check_ref
Security & privacy
Read this before pointing the server at a vault you care about.
- LLM privacy — Anything revealed to the model may be sent to your AI provider and retained under their policies.
- MCP is not end-to-end encrypted for secrets in flight — Values are plaintext inside the MCP workflow and toward the model. They are encrypted at rest in 1Password once stored.
- Best fit — Automation credentials: CI tokens, bot accounts, disposable env secrets.
- Avoid — Banking, primary personal logins, recovery codes, or anything you cannot afford to expose to a model provider.
- Token = master key — Scope the service account tightly; rotate immediately if leaked; never commit tokens or MCP configs with secrets.
- Prefer references —
op://…+op_runbeat pasting passwords into prompts or files. - Least privilege — Dedicated automation vaults beat sharing your whole account.
Protocol & compatibility
| Piece | Detail |
|---|---|
| Package | @takescake/1password-mcp |
| Runtime | Node.js ≥ 20 |
| Transport | stdio |
| MCP SDK | @modelcontextprotocol/server v2 |
| Protocol | Negotiates 2026-07-28; keeps legacy client compatibility |
| Registry name | io.github.CakeRepository/1password |
Development
git clone https://github.com/CakeRepository/1Password-MCP.git
cd 1Password-MCP
npm ci
npm run build
npm test
npm run lint
Watch mode: npm run dev.
Project layout
src/
index.ts # Entrypoint — MCP stdio + protocol negotiation
config.ts # CLI / env / Keychain / allow-list
client.ts # 1Password SDK client
logger.ts # Structured logs on stderr (stdout is protocol)
secret-ref.ts # op:// parsing & allow-list checks
utils.ts # Result helpers, password generation
tools/ # All 15 MCP tools
prompts/ # Interactive workflow prompts
resources/ # 1password:// resources
tests/
See CONTRIBUTING.md. Maintainers / agents: AGENTS.md.
Changelog
See CHANGELOG.md for version history, including the 4.0.0 MCP v2 / 2026-07-28 migration and the 3.0.0 op_run / reveal-opt-in security changes.
License
常见问题
io.github.CakeRepository/1password 是什么?
提供用于 1Password service accounts 的 MCP server,支持安全集成与凭据访问管理。
相关 Skills
前端设计
by anthropics
面向组件、页面、海报和 Web 应用开发,按鲜明视觉方向生成可直接落地的前端代码与高质感 UI,适合做 landing page、Dashboard 或美化现有界面,避开千篇一律的 AI 审美。
✎ 想把页面做得既能上线又有设计感,就用前端设计:组件到整站都能产出,难得的是能避开千篇一律的 AI 味。
网页应用测试
by anthropics
用 Playwright 为本地 Web 应用编写自动化测试,支持启动开发服务器、校验前端交互、排查 UI 异常、抓取截图与浏览器日志,适合调试动态页面和回归验证。
✎ 借助 Playwright 一站式验证本地 Web 应用前端功能,调 UI 时还能同步查看日志和截图,定位问题更快。
网页构建器
by anthropics
面向复杂 claude.ai HTML artifact 开发,快速初始化 React + Tailwind CSS + shadcn/ui 项目并打包为单文件 HTML,适合需要状态管理、路由或多组件交互的页面。
✎ 在 claude.ai 里做复杂网页 Artifact 很省心,多组件、状态和路由都能顺手搭起来,React、Tailwind 与 shadcn/ui 组合效率高、成品也更精致。
相关 MCP Server
GitHub
编辑精选by GitHub
GitHub 是 MCP 官方参考服务器,让 Claude 直接读写你的代码仓库和 Issues。
✎ 这个参考服务器解决了开发者想让 AI 安全访问 GitHub 数据的问题,适合需要自动化代码审查或 Issue 管理的团队。但注意它只是参考实现,生产环境得自己加固安全。
Context7 文档查询
编辑精选by Context7
Context7 是实时拉取最新文档和代码示例的智能助手,让你告别过时资料。
✎ 它能解决开发者查找文档时信息滞后的问题,特别适合快速上手新库或跟进更新。不过,依赖外部源可能导致偶尔的数据延迟,建议结合官方文档使用。
by tldraw
tldraw 是让 AI 助手直接在无限画布上绘图和协作的 MCP 服务器。
✎ 这解决了 AI 只能输出文本、无法视觉化协作的痛点——想象让 Claude 帮你画流程图或白板讨论。最适合需要快速原型设计或头脑风暴的开发者。不过,目前它只是个基础连接器,你得自己搭建画布应用才能发挥全部潜力。