1. Open Claude.ai → Settings → Integrations.
2. Click Add MCP Server and paste the URL above.
3. Claude.ai redirects you to faktry — sign in and click Approve.
4. Done — Claude.ai stores the token automatically.

1. Open Claude Desktop → Settings → Developer → MCP Servers.
2. Click Add Server and paste the URL above.
3. A browser window opens — sign in to faktry and click Approve.
4. Restart Claude Desktop if tools don't appear immediately.

1. Open Cursor → Settings → AI → MCP Servers.
2. Click Add MCP Server and paste the URL above.
3. Cursor detects OAuth and opens a browser login window.
4. Sign in to faktry, approve access, and return to Cursor.

1. Open Hermes Agent → Settings → MCP Servers.
2. Click + and paste the URL above.
3. Hermes opens a browser login — sign in to faktry and approve.
4. Toggle the server on — tools appear immediately.

1. Enable Developer Mode in Settings → Beta Features.
2. Go to Settings → MCP Servers → Add Server.
3. Paste the URL above and follow the browser OAuth flow.

Add to .mcp.json (project) or ~/.claude.json (user):

{
  "mcpServers": {
    "faktry": {
      "type": "http",
      "url": "https://faktry.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY_HERE" }
    }
  }
}

Picked up automatically — no restart needed.

Add to .vscode/mcp.json (workspace) or via Settings → MCP Servers:

{
  "servers": {
    "faktry": {
      "type": "http",
      "url": "https://faktry.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY_HERE" }
    }
  }
}

Edit ~/.codeium/windsurf/mcp_config.json (%USERPROFILE%\.codeium\windsurf\mcp_config.json on Windows) — note Windsurf uses serverUrl, not url:

{
  "mcpServers": {
    "faktry": {
      "serverUrl": "https://faktry.ai/api/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY_HERE" }
    }
  }
}

Restart Windsurf to activate the tools.

Add to ~/.continue/config.yaml:

mcpServers:
  faktry:
    type: http
    url: https://faktry.ai/api/mcp
    requestOptions:
      headers:
        Authorization: "Bearer YOUR_API_KEY_HERE"

Continue.dev reloads automatically.

Add to ~/.codex/config.toml:

[mcp.servers.faktry]
type = "http"
url = "https://faktry.ai/api/mcp"

[mcp.servers.faktry.headers]
Authorization = "Bearer YOUR_API_KEY_HERE"

Verify with codex mcp list.

The @faktry/mcp package is a thin stdio bridge to the hosted server. Requires Node 18+.

Generic form:

FAKTRY_API_KEY=YOUR_API_KEY_HERE npx -y @faktry/mcp

Claude Desktop (claude_desktop_config.json) and most other mcpServers clients:

{
  "mcpServers": {
    "faktry": {
      "command": "npx",
      "args": ["-y", "@faktry/mcp"],
      "env": { "FAKTRY_API_KEY": "YOUR_API_KEY_HERE" }
    }
  }
}

Zed (settings.json, %APPDATA%\Zed\settings.json on Windows):

{
  "context_servers": {
    "faktry": {
      "source": "custom",
      "command": "npx",
      "args": ["-y", "@faktry/mcp"],
      "env": { "FAKTRY_API_KEY": "YOUR_API_KEY_HERE" }
    }
  }
}

Leave FAKTRY_API_KEY out to sign in through the browser (OAuth) instead.

1. Open the Kilo Code extension settings → MCP Servers.
2. Add a new server using the same JSON snippet as Claude Code above.
3. Replace YOUR_API_KEY_HERE with your actual key.

Add a remote HTTP server pointing at https://faktry.ai/api/mcp with header:

Authorization: Bearer YOUR_API_KEY_HERE

Full per-client snippets and paths are also generated at faktry.ai/mcp.

• Claude.ai / Desktop: Settings → Integrations (or Developer → MCP Servers), confirm faktry shows as connected. Remove and re-add if it failed silently — make sure pop-ups aren't blocked so the OAuth window can open.
• Claude Code: claude mcp list should show the server.
• Codex: codex mcp list should show faktry. Start a new session after adding.
• Quick smoke test: ask your client "check my faktry credits".

• For config-file clients, confirm YOUR_API_KEY_HERE was replaced with a real, active key from faktry.ai/api/keys — keys are shown only once at creation.
• For OAuth clients, disconnect and reconnect the connector to redo the flow.

Video renders and some audio/image jobs take longer than a plain image generation. The tool returns a pending job_id immediately — keep polling faktry_job_status until it reports completed.