Miftah
A local MCP auth wrapper and credential broker for multi-account workflows, enabling profile switching, secret injection, and policy enforcement for upstream MCP servers.
README
Miftah
Wrap any MCP. Use the right account without reconnecting.
Miftah (@lubab/miftah) is a local MCP auth wrapper and credential broker for multi-account workflows. It sits between an MCP client such as Claude Desktop and an existing upstream MCP server, injects the selected profile's credentials, forwards MCP operations, and records redacted audit metadata.
Miftah is not a replacement for GitHub MCP, Sentry MCP, PostHog MCP, or any other provider MCP. It wraps them. The upstream server remains responsible for provider behavior; Miftah handles profile selection, secret injection, lifecycle, routing, policy, and redaction.
Install
npm install -g @lubab/miftah
Miftah runs locally by default. It has no cloud dependency or telemetry.
Quick start
Generate a safe template:
miftah init github --preset github --output ~/.config/miftah/github.json
Edit the profile environment to reference shell variables, then validate it:
export GITHUB_WORK_TOKEN='...'
miftah validate --config ~/.config/miftah/github.json
Run a wrapped server directly:
miftah --config ~/.config/miftah/github.json
The same process can be configured in Claude Desktop:
{
"mcpServers": {
"github": {
"command": "miftah",
"args": ["--config", "/Users/me/.config/miftah/github.json"]
}
}
}
Profiles
Profiles are named credential environments. Keep secret values outside JSON whenever possible:
{
"version": "1",
"name": "github",
"defaultProfile": "work",
"upstream": {
"transport": "stdio",
"command": "docker",
"args": ["run", "-i", "--rm", "ghcr.io/github/github-mcp-server"]
},
"profiles": {
"work": {
"description": "Work GitHub",
"env": {
"GITHUB_PERSONAL_ACCESS_TOKEN": "${GITHUB_WORK_TOKEN}"
},
"policy": "safe-write"
},
"personal": {
"env": {
"GITHUB_PERSONAL_ACCESS_TOKEN": "${GITHUB_PERSONAL_TOKEN}"
},
"policy": "readonly"
}
},
"policies": {
"readonly": {
"allowRisk": ["read"],
"denyRisk": ["write", "destructive"]
},
"safe-write": {
"allowRisk": ["read", "write"],
"denyRisk": ["destructive"],
"requireConfirmation": ["write"]
}
},
"security": {
"allowPlaintextSecrets": false,
"allowProfileSwitchingFromMcp": true,
"requireExplicitProfileForDestructive": true
}
}
Claude can call miftah_list_profiles, miftah_current_profile, miftah_use_profile, miftah_profile_info, miftah_health, miftah_validate_config, miftah_list_upstream_tools, miftah_restart_profile, and miftah_route_preview. Upstream tools are exposed unchanged unless they collide with a reserved miftah_ name.
For account bundles, define upstreams instead of upstream. Tools are exposed as <upstream>__<tool> (for example github__search_issues) and each profile can provide per-upstream environment or header overrides. See examples/multi-upstream.miftah.json.
Routing and safety
Routing can use the active profile or rules matching tool arguments:
{
"routing": {
"mode": "hybrid",
"fallback": "ask",
"rules": [
{
"name": "work-repository",
"when": { "args.repo": "my-org/work" },
"profile": "work"
}
]
}
}
When several profiles match, Miftah refuses to guess. Use explicit profile switching for write and destructive actions. Local policies can deny risky tools or return a confirmation-needed result. Provider token scopes still matter: local policy cannot make a write-capable provider token read-only.
Secret handling
Supported local references include environment variables (${NAME} and secretref:env://NAME), configured dotenv files (secretref:dotenv://NAME), and explicitly opt-in plaintext (secretref:plain://...). OS keychains and external secret CLIs are reserved extension points. Secrets are redacted from diagnostics, errors, stderr forwarding, audit entries, and tool responses.
Use miftah doctor to inspect config and upstream readiness without printing process environment values.
CLI
| Command | Purpose |
|---|---|
miftah --config <file> |
Run the STDIO MCP wrapper |
miftah serve --config <file> |
Run the wrapper explicitly |
miftah validate --config <file> |
Parse and validate JSON config |
miftah doctor --config <file> |
Validate config and report redacted runtime details |
miftah init <name> --preset <generic|github|sentry> |
Generate a template |
miftah schema |
Print the JSON Schema |
miftah list-tools --config <file> [--profile <name>] |
Discover upstream tools |
miftah test-profile --config <file> --profile <name> |
Start and initialize one profile |
miftah logs --config <file> |
Read the configured JSONL audit log |
Documentation
Current boundaries
The current release implements local STDIO and remote HTTP/SSE upstream clients, profile switching, routing rules, policies, tools/resources/prompts proxying, process/session caching, redacted audit logging, and a publishable CLI. The configuration model reserves interfaces for multi-upstream account bundles, persisted state, and optional UI.
License
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。
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。