Developer / MCP

Step 7 of 25

Connect an AI coding tool to Altified

MCP lets tools such as Cursor, Claude, and VS Code agents ask Altified about this project and perform approved actions for you.

MCP stands for Model Context Protocol. In plain language, it is a safe connection between an AI coding tool and your Altified project. The AI tool does not connect directly to your database. It talks to the Altified MCP server, which checks your token and permissions before using Altified.

The connection looks like this

Your AI tool → Altified MCP server → Your Altified project

What are MCP settings?

MCP settings are the place inside an AI code editor where you add outside tools for the AI to use. They are not a setting in the Altified dashboard. Look in your editor for a menu or file named MCP, MCP servers, Tools, or Integrations.

The editor must support MCP. A normal text editor without an AI assistant cannot use this connection.

Set it up from beginning to end

  1. Open your project. In the Altified dashboard, open the project you want your AI coding tool to use, then choose MCP from the project cards or the project navigation.
  2. Choose permissions. Pick Let my AI set things up if you want the AI to integrate Altified for you, or Read only if it should only inspect your project. Every change still asks for your confirmation.
  3. Create a token. Give it a name such as Cursor on my laptop. Tokens created from a project are limited to that project.
  4. Copy the token immediately. Altified shows the complete token only once. It works like a password for the AI tool.
  5. Add Altified to your editor. After the token is created, pick your editor in the token panel. Cursor and VS Code have an Install button; Claude Code gives you a command to run; other editors get JSON with your token already filled in.
  6. Using another editor? Search its settings for MCP, then choose an option such as Add server, New MCP server, or Edit configuration. The wording is different in each editor.
  7. Paste and save the configuration. If the editor accepts JSON, paste the configuration you copied from Altified. If it asks for separate fields, use the server URL, the Authorization header, and the token shown below.
  8. Restart or reload the editor. Some editors only connect to new MCP servers after the AI assistant or editor has been restarted. Approve or enable the Altified server if the editor asks.
  9. Ask it to integrate Altified. Open your app's folder in the editor and ask: “Integrate Altified into this project.” The AI reads your project's setup instructions, shows you a plan, waits for your approval, then installs the SDK, places the language switcher, and checks that the SDK is connected.

Run in your project's terminal.

claude mcp add --transport http altified https://mcp.altified.com/mcp \
  --header "Authorization: Bearer YOUR_MCP_TOKEN"

The examples above use YOUR_MCP_TOKEN as a placeholder. The MCP page in your project fills in your real token automatically. Never publish the completed configuration.

Other things to ask

  • “Run a health check of my Altified project.” — SDK connection, secret key, switcher, translation progress, usage, and the latest SEO audit.
  • “Help me configure the Altified language switcher.” — review languages, order, labels, and style, then publish after you approve.
  • “Which translations are missing for French?” — list missing or failed content and request translations after you approve.

Your AI will tell you to create the SDK secret key in the dashboard and paste it into .env yourself. MCP never creates or shows secret keys.

If your editor uses separate fields

Editor fieldWhat to enter
Server namealtifiedServer URLhttps://mcp.altified.com/mcpHeader nameAuthorizationHeader valueBearer YOUR_MCP_TOKEN

The token is private

Treat the MCP token like a password. Do not put it in Git, a public issue, a screenshot, frontend source code, or a NEXT_PUBLIC_ environment variable. If it is exposed, revoke it in Altified and create a replacement.

If it does not connect

  • The server does not appear: confirm that the editor supports MCP, then restart or reload it.
  • Unauthorized or 401: copy the configuration again, check that the token is complete, and make sure it has not been revoked.
  • Permission denied or 403: return to Altified and create a token with the required permission. A project token also cannot access another project.
  • The AI cannot change anything: check that the token has the required write permission and approve the requested action when Altified asks for confirmation.

What the permissions mean

  • projects:read — lets the AI see project names, domains, languages, and status.
  • languages:read — lets the AI see the languages Altified supports.
  • translations:read — lets the AI see translation progress.
  • seo:read — lets the AI read SEO audit status and findings.
  • integration:read — lets the AI read safe SDK setup information.
  • usage:read — lets the AI read plan and usage information.
  • projects:write — lets the AI request project and language switcher changes, including publishing the switcher.
  • languages:write — lets the AI request adding or removing a language.
  • translations:write — lets the AI request, retry, or edit translations, using your translation allowance.

Every write still requires your confirmation. Altified also checks project ownership, token permissions, and plan limits. seo:write is reserved for future actions.

What MCP does not do

  • It does not replace the Altified dashboard or SDK.
  • It does not give the AI your database, payment details, or backend secrets.
  • It does not automatically edit your codebase; the AI coding tool makes code changes locally.
  • It does not let a project token access another Altified project.

Revoking access

Return to the project's MCP page, find the token, and choose Revoke. The AI tool will stop being able to connect with that token. Create a new token if you lose the old one or share it accidentally.

Need the normal application integration instead? Continue with the SDK setup guide.