An MCP server directory helps you find Model Context Protocol servers for AI coding workflows, compare auth and transport requirements, and copy starter configs for Claude Code, Cursor, Codex CLI, Claude Desktop, VS Code, and other MCP clients.
Filter common MCP servers by category, auth model, transport, and use case, then copy a starter config for your AI client.
Give an AI client scoped read/write access to approved local folders.
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-filesystem",
"~/src"
]
}
}
}Pass only the folders the client should touch. Avoid broad home-directory access unless you are in a trusted session.
Inspect repositories, issues, pull requests, comments, releases, and workflow status.
{
"mcpServers": {
"github": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-github"
]
}
}
}Create a GitHub token with the narrowest repo and workflow scopes your automation needs, then expose it as GITHUB_PERSONAL_ACCESS_TOKEN.
Run read-focused SQL queries against a Postgres database from an AI client.
{
"mcpServers": {
"postgres": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-postgres",
"postgresql://user:pass@host/db"
]
}
}
}Use a read-only database role for analysis workflows. Keep production write credentials out of local MCP config.
Query a local SQLite database for app state, fixtures, exports, or analytics snapshots.
{
"mcpServers": {
"sqlite": {
"command": "npx",
"args": [
"-y",
"mcp-server-sqlite",
"--db-path",
"./data/app.db"
]
}
}
}Point the server at a copy of sensitive databases when possible so agent experiments do not mutate live local data.
Let an AI client open pages, click controls, capture screenshots, and verify browser behavior.
{
"mcpServers": {
"playwright": {
"command": "npx",
"args": [
"-y",
"@playwright/mcp@latest"
]
}
}
}Use a dedicated browser profile for automation when testing authenticated apps or destructive admin flows.
Automate Chromium for page inspection, screenshots, PDF generation, and scripted UI checks.
{
"mcpServers": {
"puppeteer": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-puppeteer"
]
}
}
}Prefer Playwright for complex test flows, and Puppeteer for simple browser automation in lightweight environments.
Search channels, summarize threads, and draft or send team messages with workspace context.
{
"mcpServers": {
"slack": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-slack"
]
}
}
}Install a Slack app with channel-scoped permissions and keep posting permissions separate from read-only analysis setups.
Find, summarize, and update documents, spreadsheets, and shared workspace files.
{
"mcpServers": {
"google-drive": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-gdrive"
]
}
}
}Run OAuth setup interactively once, then reuse the generated credentials from the same machine profile.
Inspect production errors, stack traces, releases, and issue context from Sentry.
{
"mcpServers": {
"sentry": {
"command": "npx",
"args": [
"-y",
"@sentry/mcp-server"
]
}
}
}Use a token limited to the projects and organizations your agent should inspect.
Read, create, and update issues, projects, cycles, and engineering task context.
{
"mcpServers": {
"linear": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.linear.app/sse"
]
}
}
}Remote MCP servers usually require a browser OAuth flow. Confirm the exact workspace before letting an agent write.
Search workspace pages, update databases, and turn project context into structured docs.
{
"mcpServers": {
"notion": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.notion.com/mcp"
]
}
}
}Limit access to the teamspaces and databases your automation actually needs, especially for write workflows.
Run web searches through Brave Search API for research, citations, and competitive checks.
{
"mcpServers": {
"brave-search": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-brave-search"
]
}
}
}Store BRAVE_API_KEY in the server env section and keep search-based claims cited in downstream reports.
An MCP server directory is a searchable list of Model Context Protocol servers, including what each server connects to, how it authenticates, and which AI coding workflows it supports.
Start with the server that matches your most repeated workflow. Developers usually begin with Filesystem, GitHub, Playwright, or Postgres because those servers map directly to repo work, PRs, browser checks, and database lookups.
MCP servers are safest when access is scoped. Use read-only tokens where possible, limit filesystem folders, separate production credentials from local testing, and confirm write targets before allowing an agent to update remote systems.
Stdio MCP servers run as local child processes launched by your AI client. HTTP MCP servers run remotely or separately and are reached over a network endpoint, often with OAuth or token-based authentication.
Yes. MCP servers add tools to AI clients, while Tornic turns those clients into repeatable workflows. You can use MCP-enabled Claude, Codex, or Cursor sessions inside Tornic automations.