project Claude Code config<\/a>. There is one safety gate worth knowing: the first time anyone uses a project-scoped server they are prompted to approve it, and until they do it shows as pending approval and stays disconnected. That stops a malicious repo from running a server on your machine the moment you clone it. If you approve by accident or want the prompt back, reset the choices:<\/p>\n\n\nclaude mcp reset-project-choices<\/code><\/pre>\n\n\nWith the wiring understood, the next part is credentials.<\/p>\n\n
Pass tokens and environment variables<\/h2>\n\n Real servers need secrets: a GitHub token, a database URL, an API key. Two flags carry them, -e<\/code> for environment variables and -H<\/code> (or --header<\/code>) for HTTP headers. Both are variadic, meaning they swallow every value that follows. So they must come after<\/strong> the name and URL, never before.<\/p>\n\nPut a flag like --header<\/code> ahead of the name and the parser eats the name, leaving you with error: missing required argument 'name'<\/code>. The order that works for a token-protected HTTP server:<\/p>\n\n\nclaude mcp add --scope project --transport http github https:\/\/api.githubcopilot.com\/mcp\/ --header \"Authorization: Bearer YOUR_GITHUB_PAT\"<\/code><\/pre>\n\n\nName and URL first, header last. The confirmation redacts the secret so it never lands in your scrollback, and it trims the trailing slash from the echoed URL even though the stored .mcp.json<\/code> keeps it:<\/p>\n\n\nAdded HTTP MCP server github with URL: https:\/\/api.githubcopilot.com\/mcp\nHeaders: {\n \"Authorization\": \"[REDACTED]\"\n}<\/code><\/pre>\n\n\nFor a stdio server that reads its config from the environment, the same rule holds. The -e<\/code> pairs go after the name, before the --<\/code>:<\/p>\n\n\nclaude mcp add postgres -e DATABASE_URL=postgresql:\/\/localhost\/mydb -- npx -y @modelcontextprotocol\/server-postgres<\/code><\/pre>\n\n\nHardcoding secrets into a shared .mcp.json<\/code> is a bad idea, so the config supports expansion. Reference an environment variable with ${VAR}<\/code>, and fall back to a default with ${VAR:-default}<\/code>. Expansion works in command<\/code>, args<\/code>, env<\/code>, url<\/code>, and headers<\/code>:<\/p>\n\n\n{\n \"mcpServers\": {\n \"github\": {\n \"type\": \"http\",\n \"url\": \"https:\/\/api.githubcopilot.com\/mcp\/\",\n \"headers\": { \"Authorization\": \"Bearer ${GITHUB_PAT}\" }\n }\n }\n}<\/code><\/pre>\n\n\nNow the repo carries the shape and each developer supplies their own GITHUB_PAT<\/code>.<\/p>\n\nCheck and manage your servers<\/h2>\n\n One command lists everything across all scopes and health-checks each one as it goes:<\/p>\n\n\n
claude mcp list<\/code><\/pre>\n\n\nConnected servers show a check, ones still waiting on sign-in say so plainly:<\/p>\n\n\n
Checking MCP server health...\n\ngithub: https:\/\/api.githubcopilot.com\/mcp\/ (HTTP) - \u2713 Connected\nplaywright: npx -y @playwright\/mcp - \u2713 Connected\nfilesystem: npx -y @modelcontextprotocol\/server-filesystem ~\/projects - \u2713 Connected\nsentry: https:\/\/mcp.sentry.dev\/mcp (HTTP) - ! Needs authentication<\/code><\/pre>\n\n\nFor one server’s details, including its scope and status, use get<\/code>:<\/p>\n\n\nclaude mcp get sentry<\/code><\/pre>\n\n\nThat same status view appears when you run the list command, shown here against a live set of servers:<\/p>\n\n\nA server that says it needs authentication is waiting on you. Open the panel from inside a session:<\/p>\n\n\n
\/mcp<\/code><\/pre>\n\n\nThe panel lists every server, the tools each one exposes, and an Authenticate option for the ones that need it. Pick the server, sign in through the browser window it opens, and Claude reconnects on its own. To drop a server, name it and the scope it lives in:<\/p>\n\n\n
claude mcp remove sentry -s project<\/code><\/pre>\n\n\nThe server leaves the shared config and the file updates in place.<\/p>\n\n
Use the tools Claude just gained<\/h2>\n\n Connecting a server is half the job. Most guides stop at the green check, but a connected server gives you three concrete handles inside a session, and they are where the real speedup lives.<\/p>\n\n
That same \/mcp<\/code> panel lists the exact tools each server exposes, so you can see what Claude can now call. Many servers also ship prompts<\/em>, which surface as slash commands named after the server. Type \/<\/code> and they show up alongside the built-ins:<\/p>\n\n\n\/mcp__github__list_prs\n\/mcp__jira__create_issue \"Bug in login flow\" high<\/code><\/pre>\n\n\nServers can also expose resources<\/em>, which you pull into a prompt with an @<\/code> mention the same way you reference a file. The format is @server:protocol:\/\/resource\/path<\/code>:<\/p>\n\n\nCan you analyze @github:issue:\/\/123 and suggest a fix?<\/code><\/pre>\n\n\nThe last handle is permissions. By default Claude asks before each new MCP tool call, which gets old fast for a server you trust. MCP tools are named mcp__<server>__<tool><\/code>, and you pre-approve a whole server by naming it bare. Drop this into .claude\/settings.json<\/code> and the GitHub server’s tools run without a prompt:<\/p>\n\n\n{\n \"permissions\": {\n \"allow\": [\"mcp__github\"]\n }\n}<\/code><\/pre>\n\n\nThe wildcard form mcp__github__*<\/code> does the same thing, and mcp__github__get_*<\/code> narrows it to just the read tools. For a one-off session, pass --allowedTools \"mcp__github\"<\/code> on launch instead of editing settings. One more direction worth knowing: Claude Code can run as<\/em> a server itself with claude mcp serve<\/code>, exposing its own tools to another MCP client like Claude Desktop or an IDE. Niche, but it is how you slot Claude Code into a larger multi-agent setup.<\/p>\n\nBest MCP servers to start with<\/h2>\n\n You do not need a dozen. A handful of well-chosen servers covers most real work, and each one you add ships more tool definitions into the context window. Start here:<\/p>\n\n\nServer<\/th> What it adds<\/th> Transport<\/th><\/tr><\/thead> \nGitHub<\/td> Issues, pull requests, code search, reviews<\/td> HTTP (hosted)<\/td><\/tr>\n Playwright<\/td> Drives a real browser for testing and scraping<\/td> stdio<\/td><\/tr>\n Context7<\/td> Version-pinned docs for thousands of libraries<\/td> stdio<\/td><\/tr>\n Postgres or SQLite<\/td> Query and inspect your database directly<\/td> stdio<\/td><\/tr>\n Sentry<\/td> Pulls live errors and traces into the session<\/td> HTTP (hosted)<\/td><\/tr>\n Sequential Thinking<\/td> Structured, step-by-step reasoning<\/td> stdio<\/td><\/tr>\n<\/tbody><\/table><\/figure>\n\n\nThree to six is plenty. Past that, the tool list grows, startup slows, and you spend context you could give the actual task. If your sessions feel heavy, the same discipline that keeps token usage down<\/a> applies here: trim servers you are not using. The GitHub and database servers alone change how much Claude Code handles for DevOps work<\/a>, from triaging issues to reading production schemas.<\/p>\n\nTroubleshooting the common failures<\/h2>\n\n These are the four that actually send people looking for answers, with the fix for each.<\/p>\n\n
The add command says error: missing required argument 'name'<\/code>.<\/strong> A variadic flag like --header<\/code> or -e<\/code> ran ahead of the server name and ate it. Move every flag after the name and URL, and put the stdio command after --<\/code>.<\/p>\n\nA project server is stuck on pending approval.<\/strong> This trips people up the first time they clone a repo with a .mcp.json<\/code>. Servers from that file are disabled until you approve them, by design. Run \/mcp<\/code> inside a session to approve, or claude mcp reset-project-choices<\/code> if you need the prompt back.<\/p>\n\nA server times out while starting.<\/strong> Slow stdio servers (a cold npx<\/code> download, a heavy runtime) can blow the startup window. Raise it for the session, or pin a per-server limit in .mcp.json<\/code>:<\/p>\n\n\nMCP_TIMEOUT=15000 claude<\/code><\/pre>\n\n\nFor a per-server cap, add a \"timeout\"<\/code> field in milliseconds to that server’s entry, which overrides the global MCP_TOOL_TIMEOUT<\/code> for that one server.<\/p>\n\nClaude warns that MCP tool output exceeds 10,000 tokens.<\/strong> That is the warning line; the hard cap defaults to 25,000 tokens, and a chatty server (a wide database query, a giant API response) burns context you wanted for the task. Raise the ceiling when you genuinely need the data:<\/p>\n\n\nMAX_MCP_OUTPUT_TOKENS=50000 claude<\/code><\/pre>\n\n\nThe gotcha here is that more output is rarely the real fix. Trimming a noisy server or narrowing the query usually beats raising the cap. When you are properly stuck, launch with claude --debug<\/code> and it prints the underlying MCP server errors instead of a quiet failure.<\/p>\n\nMCP command cheat sheet<\/h2>\n\n Everything you need to run, in one place:<\/p>\n\n\nTask<\/th> Command<\/th><\/tr><\/thead> \nAdd stdio server<\/td> claude mcp add -s project NAME -- CMD ARGS<\/code><\/td><\/tr>\nAdd HTTP server<\/td> claude mcp add -s project -t http NAME URL<\/code><\/td><\/tr>\nAdd with token<\/td> claude mcp add ... NAME URL --header \"Authorization: Bearer TOKEN\"<\/code><\/td><\/tr>\nList and health-check<\/td> claude mcp list<\/code><\/td><\/tr>\nInspect one server<\/td> claude mcp get NAME<\/code><\/td><\/tr>\nRemove a server<\/td> claude mcp remove NAME -s project<\/code><\/td><\/tr>\nAuthenticate or reconnect<\/td> \/mcp<\/code> (inside a session)<\/td><\/tr>\nRun a server prompt<\/td> \/mcp__SERVER__PROMPT args<\/code><\/td><\/tr>\nAllow a server’s tools<\/td> --allowedTools \"mcp__SERVER\"<\/code> or permissions.allow<\/code><\/td><\/tr>\nRaise output cap<\/td> MAX_MCP_OUTPUT_TOKENS=50000 claude<\/code><\/td><\/tr>\n<\/tbody><\/table><\/figure>\n\n\nPin this next to your Claude Code cheat sheet<\/a> and you have the whole MCP surface covered. Add the servers your stack actually uses, commit the .mcp.json<\/code>, and the next clone is wired up on first run.<\/p>","protected":false},"excerpt":{"rendered":"Claude Code is sharp on its own. Point it at MCP servers and it can read your database, drive a browser, open GitHub issues, and pull current library docs without you pasting anything into the prompt. The model gains tools, and it uses them on its own. This is a complete Claude Code MCP setup. … Read more<\/a><\/p>\n","protected":false},"author":27,"featured_media":168492,"comment_status":"open","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[39034,690,299],"tags":[17245,212,669,211],"cfg_series":[39880],"class_list":["post-168493","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","category-dev","category-how-to","tag-ai","tag-automation","tag-dev","tag-devops","cfg_series-claude-code-mastery"],"_links":{"self":[{"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/posts\/168493","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/users\/27"}],"replies":[{"embeddable":true,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/comments?post=168493"}],"version-history":[{"count":3,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/posts\/168493\/revisions"}],"predecessor-version":[{"id":168685,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/posts\/168493\/revisions\/168685"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/media\/168492"}],"wp:attachment":[{"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/media?parent=168493"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/categories?post=168493"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/tags?post=168493"},{"taxonomy":"cfg_series","embeddable":true,"href":"https:\/\/computingforgeeks.com\/wp-json\/wp\/v2\/cfg_series?post=168493"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"Adding](\"https:\/\/computingforgeeks.com\/wp-content\/uploads\/2026\/06\/wm-claude-code-mcp-add-server-config.png\" "\\"\\"")
<\/figure>\n\n\n![\"claude](\"https:\/\/computingforgeeks.com\/wp-content\/uploads\/2026\/06\/wm-claude-code-mcp-list-server-status.png\" "\\"\\"")
<\/figure>\n\n\n