Connect your AIin five minutes

You do not need a developer for this. Pick your AI tool below and follow a few short steps. If you can copy and paste, you can do it alone.

Setup
Show me my Proginter account profile.
Profile loaded - account active, billing on file.

Pick your tool

MCP (Model Context Protocol) is the standard AI tools use to connect to a service - Proginter speaks it. Each guide below is complete on its own - open yours and follow it top to bottom. Every path ends the same way: your AI managing your hosting for you.

Connect Claude

Works on claude.ai and in Claude Desktop, on every Claude plan - the Free plan includes one custom connector. One address, one sign-in, about two minutes. The sign-in path needs no API key and no config file. Prefer a key? Use the API key tab.

74sSilent walkthrough with example data on screen.

Sign in

  1. Open Connectors

    In Claude, open Customize → Connectors, click the + button, and choose Add custom connector. On a Team or Enterprise plan, an owner adds it under Organization settings → Connectors first.

  2. Paste the Proginter address

    Name it Proginter and paste this address. Leave every other field empty, then click Add.

    https://proginter.com/mcp
  3. Sign in and approve

    Claude opens a Proginter sign-in page in your browser. Check the address bar shows proginter.com, sign in, review the access it asks for, and click Approve. If the page does not open on its own, copy the link Claude shows and open it yourself.

  4. Try it

    Back in Claude, ask "Show me my Proginter account profile." A reply with your name and email means you are connected. If Claude does not use Proginter, click the + button in the chat, hover over Connectors, and turn it on for that conversation.

API key

Claude Desktop can also connect through a config file instead of the sign-in flow. Create an API key in your panel under Developer Tools → API - choose the MCP (AI assistant) option, the secret is shown once. Then add this to your claude_desktop_config.json with your real key and secret and restart Claude Desktop. The file stores the secret in plain text - anything running under your user account can read it, so keep the key's permissions small and revoke it when you are done. Then ask "Show me my Proginter account profile." - your name and email back means it works, a 401 means a wrong key or a missing header. It uses the mcp-remote bridge - the plain url config does not work in Claude Desktop:

{
  "mcpServers": {
    "proginter": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@0.1.38",
        "https://proginter.com/mcp",
        "--header",
        "X-API-Key:${PROGINTER_KEY}",
        "--header",
        "X-API-Secret:${PROGINTER_SECRET}"
      ],
      "env": {
        "PROGINTER_KEY": "pk_your_public_key",
        "PROGINTER_SECRET": "sk_your_secret_key"
      }
    }
  }
}

Connect ChatGPT

Works on chatgpt.com for Plus, Pro, Business, Enterprise, and Education plans. OpenAI rolls this out in stages - on some plans the connector is read-only for now. You set it up once in your browser - there is no API key to manage. The ChatGPT desktop app has no setup screen for this - do it at chatgpt.com.

82sSilent walkthrough with example data on screen.

Connector

  1. Turn on Developer mode

    In ChatGPT, open Settings → Security and login and turn on Developer mode. You only do this once. On a Business, Enterprise, or Education workspace, an admin may need to allow it first. If you do not see it, your plan does not include it - the Custom GPT method on this page is an alternative when you have GPT-builder access (a paid ChatGPT feature). Otherwise connect through Claude, Cursor, or VS Code instead.

  2. Add the Proginter connector

    Open ChatGPT Plugins, click the + button, and create an app for the Proginter server. Name it Proginter, paste the address below, choose OAuth as the authentication, and create it. The new app appears under Drafts.

    https://proginter.com/mcp
  3. Sign in and approve

    ChatGPT sends you to a Proginter sign-in page - right after you create the app, or the first time it runs in a chat. Check the address bar shows proginter.com, sign in, review the access it asks for, and click Approve.

  4. Try it

    In a chat, open the plus menu, choose Developer mode, and select Proginter for the conversation. Then ask "Show me my Proginter account profile." A reply with your name and email means you are connected.

Custom GPT

No Developer mode on your plan, or building your own GPT? A Custom GPT connects through OAuth credentials instead of a connector. Building GPTs needs a paid ChatGPT plan.

  1. Get the callback URL

    In the GPT builder, add an Action and choose Authentication: OAuth. The builder shows you a callback URL - copy it.

  2. Register the app in your panel

    In your panel, open OAuth AppsRegister App. Name it after your own project (names that impersonate Proginter are rejected) and paste the callback URL as the redirect URL. A new app starts unverified, so pick only the Profile permission for the first test. Until verification, every permission except Profile and Affiliate makes the sign-in refuse instead of granting less - and Affiliate includes payout settings, so add it only on purpose. Every new app enters the staff review queue automatically - to speed one up, write to support@proginter.com. Save. You get a client_id and a client_secret - the secret is shown once, so copy both, plus the scope string the same screen shows.

  3. Fill in the GPT builder

    Back in the builder, paste the client_id and client_secret, set the authorization URL to https://proginter.com/client/oauth/authorize, the token URL to https://proginter.com/oauth/token, and the same permissions space-separated in the Scope field (for a Profile-only first test that is account:profile). Leave Token Exchange Method on the default (POST request).

  4. Upload the schema and try it

    Import the schema from https://proginter.com/api/v1/openapi.json, save the GPT, and ask it "Show me my Proginter account profile." A reply with your name and email means you are connected.

Connect Cursor

Cursor signs in with your Proginter account - one click, no API key to copy. Prefer a key for automation? That works too. About two minutes.

72sSilent walkthrough with example data on screen.

Sign in

  1. Add Proginter to Cursor

    Click the button below - Cursor opens with the Proginter server already filled in. Click Install. You can also add it yourself in Cursor under Customize → MCP with this address:

    https://proginter.com/mcp
    Add to Cursor
  2. Sign in

    Cursor asks you to log in to the new server. A Proginter sign-in page opens in your browser - check the address bar shows proginter.com, sign in, review the access it asks for, and click Approve.

  3. Try it

    Ask Cursor "Show me my Proginter account profile." A reply with your name and email means you are connected.

API key

Good for automation, or when the sign-in is not available. In your panel, open Developer Tools → API and click Create API Key - choose the MCP (AI assistant) option, give it only the permissions you want the AI to have, and copy both values (the secret is shown once, and you can revoke the key on the same page at any time). Then add this to ~/.cursor/mcp.json and restart Cursor. The file stores the secret in plain text - anything running under your user account can read it, so keep the key's permissions small and revoke it when you are done. The panel's Connect your AI assistant card offers the same key ready to paste. That card uses the bridge form from the Any other tool section - it works too. Then ask "Show me my Proginter account profile." - your name and email back means it works, a 401 means a wrong key or a missing header.

{
  "mcpServers": {
    "proginter": {
      "url": "https://proginter.com/mcp",
      "headers": {
        "X-API-Key": "pk_your_public_key",
        "X-API-Secret": "sk_your_secret_key"
      }
    }
  }
}

Older Cursor that does not accept a url entry? Use this instead:

{
  "mcpServers": {
    "proginter": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@0.1.38",
        "https://proginter.com/mcp",
        "--header",
        "X-API-Key:${PROGINTER_KEY}",
        "--header",
        "X-API-Secret:${PROGINTER_SECRET}"
      ],
      "env": {
        "PROGINTER_KEY": "pk_your_public_key",
        "PROGINTER_SECRET": "sk_your_secret_key"
      }
    }
  }
}

Connect Claude Code

One command in your terminal, then sign in with your Proginter account. The sign-in path needs no API key. Prefer a key? Use the API key tab.

76sSilent walkthrough with example data on screen.

Sign in

  1. Add the server

    Run this in your terminal. The --scope user part makes Proginter available in all your projects, not just the current folder:

    claude mcp add --transport http proginter --scope user https://proginter.com/mcp
  2. Sign in

    Start Claude Code, type /mcp, pick proginter and authenticate. A browser window opens - sign in to Proginter, review the access it asks for, and approve. To check the connection any time, run claude mcp list - proginter should show Connected.

  3. Try it

    Ask "Show me my Proginter account profile." A reply with your name and email means you are connected.

API key

Good for servers and automation. Create a key in your panel under Developer Tools → API, then add it like this. The command lands in your shell history - on a shared machine prefer the sign-in path:

claude mcp add --transport http proginter --scope user \
  https://proginter.com/mcp \
  --header "X-API-Key: pk_your_public_key" \
  --header "X-API-Secret: sk_your_secret_key"

Connect VS Code

For GitHub Copilot agent mode, on any Copilot plan. One palette command, about two minutes.

80sSilent walkthrough with example data on screen.

Sign in

  1. Add the server

    Open the Command Palette (Ctrl+Shift+P, on Mac Cmd+Shift+P), run MCP: Add Server, choose HTTP, and paste this address. VS Code asks for a Server ID - type proginter. When it asks where to save, choose Global so it works in every project:

    https://proginter.com/mcp
  2. Allow and sign in

    VS Code asks you to allow the new server - approve it, then sign in to Proginter when the browser opens and review the access it asks for before you approve.

  3. Try it

    In Copilot Chat, switch to agent mode and ask "Show me my Proginter account profile." A reply with your name and email means you are connected.

API key

No sign-in needed with a key. Create one in your panel under Developer Tools → API, then put this in your user-level MCP config (Command Palette → MCP: Open User Configuration). VS Code asks for the key and secret the first time the server starts and keeps them out of the file. Then ask "Show me my Proginter account profile." - your name and email back means it works, a 401 means a wrong key or a missing header:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "proginter-key",
      "description": "Proginter API key (pk_...)"
    },
    {
      "type": "promptString",
      "id": "proginter-secret",
      "description": "Proginter API secret (sk_...)",
      "password": true
    }
  ],
  "servers": {
    "proginter": {
      "type": "http",
      "url": "https://proginter.com/mcp",
      "headers": {
        "X-API-Key": "${input:proginter-key}",
        "X-API-Secret": "${input:proginter-secret}"
      }
    }
  }
}

Any other tool or script

Everything an MCP client or a script needs, in one place. Steps 1 and 2 give you the address and the keys. Steps 3 and 4 are two config forms, not a sequence - use whichever one your tool accepts.

  1. The address

    One MCP endpoint over Streamable HTTP. Sign-in (OAuth 2.1) and API-key headers both work:

    https://proginter.com/mcp
  2. API keys

    Create keys in your panel under Developer Tools → API. Send them as X-API-Key and X-API-Secret headers on every call. The secret is shown once - copy it before closing.

  3. Generic MCP config (clients that accept a url)

    {
      "mcpServers": {
        "proginter": {
          "url": "https://proginter.com/mcp",
          "headers": {
            "X-API-Key": "pk_your_public_key",
            "X-API-Secret": "sk_your_secret_key"
          }
        }
      }
    }
  4. Tool only speaks stdio? Use the mcp-remote bridge

    {
      "mcpServers": {
        "proginter": {
          "command": "npx",
          "args": [
            "-y",
            "mcp-remote@0.1.38",
            "https://proginter.com/mcp",
            "--header",
            "X-API-Key:${PROGINTER_KEY}",
            "--header",
            "X-API-Secret:${PROGINTER_SECRET}"
          ],
          "env": {
            "PROGINTER_KEY": "pk_your_public_key",
            "PROGINTER_SECRET": "sk_your_secret_key"
          }
        }
      }
    }

    The bridge runs with npx, so the machine needs Node.js 18 or newer - get it free at nodejs.org. This form also works as claude_desktop_config.json for Claude Desktop. The env block keeps the secret out of process lists, but the file itself stores it in plain text - keep the key's permissions small and revoke it when you are done.

Prefer plain REST?
The same actions exist as a REST API. OpenAPI schema: proginter.com/api/v1/openapi.json. Full tool catalog: proginter.com/mcp-tools-catalog. To see exactly which headers reach us: proginter.com/mcp-diagnostic/echo-headers.

Shortcut: let the AI set itself up

Already talking to an AI that can browse the web? Paste this prompt and it will read this page and walk you through your exact setup. Only use it with the proginter.com address - never let an AI follow setup steps it found on another site. And never paste your API secret into the chat - the AI should show you where the secret goes, and you put it there yourself:

Connect my Proginter hosting to this app. Read https://proginter.com/ai-access-setup/ and follow the steps for the tool we are in right now. The page has one guide per tool - if you cannot see them all, open the plain-text guide at the bottom of the page.
The whole guide as plain text

For AI assistants, and for people who like it compact. Copy this block and give it to any AI - it holds everything needed to connect.

PROGINTER MCP - INSTALL GUIDE (for AI assistants and humans)

The only Proginter MCP address is https://proginter.com/mcp.
We never ask you to add a different one.
Never send your API key or secret to any host other than proginter.com.
Never paste your API secret into a chat message - an AI should show you
where the secret goes, and you paste it into the file or prompt yourself.

Endpoint: https://proginter.com/mcp  (MCP over Streamable HTTP)
Auth: OAuth 2.1 sign-in (no key needed) OR headers X-API-Key + X-API-Secret
Create API keys: https://proginter.com/client/api-documentation
  (panel: Developer Tools > API > Create API Key > choose the MCP option,
   the secret is shown once, give it only the permissions the AI needs)
Revoke access: sign-in grants at https://proginter.com/client/account-access,
  API keys at https://proginter.com/client/api-documentation
Tool catalog: https://proginter.com/mcp-tools-catalog
Header check: https://proginter.com/mcp-diagnostic/echo-headers
  (call it WITH your headers - a plain browser click sends none of them.
   Dummy values are fine - the check reports presence only. Never put a
   real secret on a command line or in a chat for this check.)
REST alternative, OpenAPI: https://proginter.com/api/v1/openapi.json
Sign-in flows below: before approving, check the browser address bar shows
  proginter.com and read the requested access list on the consent screen.

CLAUDE (claude.ai, works in Claude Desktop and mobile too - every plan, Free includes one custom connector)
1. Customize > Connectors > click + > Add custom connector
   (Team/Enterprise: an owner adds it under Organization settings > Connectors)
2. Name: Proginter. URL: https://proginter.com/mcp. Leave other fields empty. Add.
3. Sign in to Proginter in the browser window and click Approve.
If Claude does not use it in a chat: + button in the chat > Connectors > turn Proginter on.
Claude Desktop config file: the plain url form does NOT work there - use the
mcp-remote bridge from ANY OTHER MCP CLIENT below in claude_desktop_config.json.

CLAUDE CODE (terminal)
claude mcp add --transport http proginter --scope user https://proginter.com/mcp
  (--scope user = available in all your projects, not just the current folder)
Then run /mcp inside Claude Code and authenticate.
Check: claude mcp list should show proginter as Connected.
With an API key instead (keys typed on the command line land in shell history -
prefer the sign-in form above on shared machines):
claude mcp add --transport http proginter --scope user https://proginter.com/mcp \
  --header "X-API-Key: pk_..." --header "X-API-Secret: sk_..."

CHATGPT (chatgpt.com - Plus, Pro, Business, Enterprise, Education)
  (staged rollout - on some plans the connector is read-only for now)
1. Settings > Security and login > turn on Developer mode.
2. Open https://chatgpt.com/plugins > + > create an app.
   Name: Proginter. URL: https://proginter.com/mcp. Auth: OAuth. It appears under Drafts.
3. Sign in to Proginter and approve - the sign-in page opens right after creation,
   or on the app's first run in a chat.
4. In a chat: plus menu > Developer mode > select Proginter for the conversation.

CHATGPT CUSTOM GPT (building a GPT, or no Developer mode - needs paid GPT-builder access)
1. GPT builder > add an Action > Authentication: OAuth. Copy the callback URL it shows.
2. Panel: https://proginter.com/client/oauth-clients > Register App.
   Name it after your own project - Proginter-impersonating names are rejected.
   Redirect URL = the callback URL. A new self-registered app starts unverified -
   pick ONLY the Profile permission for the first test. Until verification,
   every permission except Profile and Affiliate makes the sign-in refuse
   entirely (it does not grant a smaller set), and Affiliate includes payout
   settings - add it only on purpose. Every new app enters the staff review
   queue automatically - to speed one up, write to support@proginter.com.
   Save. Copy client_id + client_secret + the scope string on the same screen.
3. In the builder: authorization URL https://proginter.com/client/oauth/authorize,
   token URL https://proginter.com/oauth/token, same permissions space-separated in
   Scope (Profile-only = account:profile). Leave Token Exchange Method on the
   default (POST request).
4. Import the schema from https://proginter.com/api/v1/openapi.json and save.

CURSOR (sign-in, no key needed)
1. One-click install link (opens Cursor with the server filled in):
   cursor://anysphere.cursor-deeplink/mcp/install?name=proginter&config=eyJ1cmwiOiJodHRwczovL3Byb2dpbnRlci5jb20vbWNwIn0=
   Or add it yourself: Customize > MCP > url https://proginter.com/mcp
2. Cursor asks you to log in - sign in to Proginter in the browser and approve.
API key alternative: ~/.cursor/mcp.json:
   {"mcpServers":{"proginter":{"url":"https://proginter.com/mcp",
    "headers":{"X-API-Key":"pk_...","X-API-Secret":"sk_..."}}}}
   then restart Cursor. The file stores the secret in plain text -
   keep the key's permissions small.

VS CODE (GitHub Copilot agent mode)
1. Command Palette > MCP: Add Server > HTTP > https://proginter.com/mcp
   It asks for a Server ID - type proginter. Then choose Global so it
   works in every project.
   With API keys, use the USER-level mcp.json instead (Command Palette >
   MCP: Open User Configuration) - this form makes VS Code prompt for the
   key and secret at first start and keeps them out of the file:
   {"inputs":[
     {"type":"promptString","id":"proginter-key","description":"Proginter API key (pk_...)"},
     {"type":"promptString","id":"proginter-secret","description":"Proginter API secret (sk_...)","password":true}],
    "servers":{"proginter":{"type":"http","url":"https://proginter.com/mcp",
     "headers":{"X-API-Key":"${input:proginter-key}","X-API-Secret":"${input:proginter-secret}"}}}}
   Never store literal keys in a repo-tracked .vscode folder.
2. Allow the server, sign in to Proginter, or use the key headers.

ANY OTHER MCP CLIENT
Native HTTP: url https://proginter.com/mcp plus the two headers above.
Stdio-only clients (incl. claude_desktop_config.json) use the mcp-remote
bridge. It runs with npx, so the machine needs Node.js 18 or newer
(nodejs.org). The config-file form - keys stay in env values, out of
process lists (the file itself still stores the secret in plain text,
so keep the key's permissions small):
  {"mcpServers":{"proginter":{"command":"npx","args":["-y","mcp-remote@0.1.38",
   "https://proginter.com/mcp","--header","X-API-Key:${PROGINTER_KEY}",
   "--header","X-API-Secret:${PROGINTER_SECRET}"],
   "env":{"PROGINTER_KEY":"pk_...","PROGINTER_SECRET":"sk_..."}}}}
Shell-only alternative: npx mcp-remote@0.1.38 https://proginter.com/mcp \
  --header "X-API-Key:pk_..." --header "X-API-Secret:sk_..."
  (command-line keys land in shell history and process lists -
   prefer the config-file form on shared machines)

VERIFY
Ask: "Show me my Proginter account profile." Expect your name and email.
401 means wrong or missing credentials.
Destructive actions need _confirm: true in the tool call - the AI tool
sets that flag itself, so it should ask its human before setting it.
The flag exists on the MCP connection only - REST callers (scripts,
Custom GPTs) have no flag, permissions are their only gate.
The credential's permissions are the hard limit on what the AI can do.
API key: you pick those permissions when creating the key. Sign-in: the
AI tool requests a set and you approve or refuse it as a whole - read it
on the consent screen. Revoke: https://proginter.com/client/account-access

FAQ

Short answers to the things that usually go wrong.

What is MCP?

MCP (Model Context Protocol) is the standard AI tools use to connect to a service. Proginter speaks it, so tools like Claude, ChatGPT, Cursor and VS Code can manage your hosting for you - with only the access you approve.

Which connection method should I use?

Pick by your tool - all paths reach the same actions. Claude, ChatGPT, Cursor, Claude Code and VS Code: sign in (OAuth) - one click, nothing to manage. The tool asks for a permission set and you approve or refuse it as a whole, so read the list before you approve. Scripts and automation: X-API-Key headers - you pick the exact permissions when you create the key. No-code platforms (Zapier, Make.com, n8n): use the API key headers when the platform lets you set request headers - nothing to register. If it only accepts a Login button, use OAuth - a self-registered app starts unverified, so expect the Profile-only limit until Proginter reviews it. Both stay supported - we are not removing either one.

I do not see connectors in ChatGPT - where are they?

Connectors need Developer mode, available on Plus, Pro, Business, Enterprise, and Education plans: open Settings → Security and login and turn it on. If your plan does not show it, the Custom GPT method on this page works too - it needs GPT-builder access, a paid ChatGPT feature. On a free plan, connect through Claude, Cursor or VS Code instead - they work on any plan. One difference in the Custom GPT path - a self-registered app starts unverified, so pick only the Profile permission until Proginter verifies it. Every permission except Profile and Affiliate makes the sign-in refuse (details in the Custom GPT tab).

I get a 401 - what now?

Your credentials are wrong, expired, or a header is missing. X-API-Key path: check both X-API-Key and X-API-Secret are set. If it still fails, rotate the key from the API Documentation page. OAuth path: the access token may have expired after an hour - your tool should refresh it on its own. If it does not, sign in and approve again.

Does Proginter work as a Claude connector?

Yes. Add Proginter as a custom connector in Claude with just the web address - open Customize → Connectors, click +, choose Add custom connector, paste https://proginter.com/mcp, then sign in and approve. No API key or config file is needed for this path. If you prefer an API key, the mcp-remote bridge in the Any other tool section still works and your tools appear alongside your other MCP tools.

How do I revoke access I previously granted?

For sign-in (OAuth) access: open your profile menu in the panel and choose Connected Devices & Apps, or go straight to proginter.com/client/account-access. It lists every grant with its permissions and last use - click Disconnect to revoke it immediately. For API keys: revoke or rotate them under Developer Tools → API.

A CDN or proxy is between me and Proginter - what then?

Some edges strip custom X-* headers. Two fixes: (a) allow X-API-Key and X-API-Secret to pass through your CDN, or (b) use the mcp-remote bridge from the Any other tool section - the request then starts from your own machine instead of the AI vendor's cloud, which avoids stripping on the vendor's side. A proxy on your own network can still strip them - the header test below shows what actually arrives. The OAuth Bearer header is standard so most CDNs leave it alone.

How do I see what my client is actually sending?

Call https://proginter.com/mcp-diagnostic/echo-headers WITH your headers, from your own terminal: curl -s https://proginter.com/mcp-diagnostic/echo-headers -H 'X-API-Key: test' -H 'X-API-Secret: test'. Dummy values are fine - the response lists every header we received, key headers always as [redacted] and other uncommon ones as [present], so the answer you want is whether your header arrived at all. Never put a real secret on a command line or into an AI chat. Just clicking the link only shows what your browser sends, so a plain click proves nothing about your AI tool's headers.

What does "destructive confirmation" mean?

Anything that deletes, charges, or replaces data is refused unless the call carries explicit approval (_confirm: true). Most AI tools show you a prompt and set it after you say yes - that prompt is your tool's behavior, so the permissions you give the AI are the hard limit. Read-only actions never need it. The flag exists on the MCP connection - a Custom GPT or a script calls the REST API directly, where your key or app permissions are the only limit.

All connected?

Hundreds of panel actions are now one ask away - the tool catalog lists exactly what is covered. Start with something small - a backup, a log check - and go from there.