Connect your agent
https://mcp.modelfy.ai/mcp
Every client needs only that URL. How you sign in:
- OAuth (recommended): the client opens the Modelfy web, you sign in once and allow it. Nothing to copy, and you are not asked again. See Keys and sign-in.
- API key: for clients without OAuth, and for bots. Send
Authorization: Bearer mfy_live_…(or a freemfy_test_…sandbox key). See Keys and sign-in.
To check it works, ask your agent: "What's my Modelfy balance?" It should call get_account.
Claude Code
The usual way
claude mcp add --transport http --scope user modelfy https://mcp.modelfy.ai/mcp
Then, inside Claude Code, run /mcp, pick modelfy and sign in in the browser. From a shell, claude mcp login modelfy does the same. Tokens refresh on their own.
--scope user makes it available in all your projects. Leave it out to add it to the current project only, or use --scope project to write a .mcp.json your team shares (each person signs in with their own account).
With an API key instead of OAuth:
claude mcp add --transport http --scope user modelfy https://mcp.modelfy.ai/mcp \
--header "Authorization: Bearer mfy_live_…"
Optional: the plugin
The plugin brings the same server plus a skill that teaches Claude the whole photo workflow: what to ask the person, when to ask before spending, how to upload local files, and what to do after each error.
/plugin marketplace add brookmind/modelfy-mcp
/plugin install modelfy@modelfy
The plugin includes the server, so you do not need claude mcp add too (if you already added it, claude mcp remove modelfy avoids two copies). In /mcp it shows as plugin:modelfy:modelfy; sign in there.
claude.ai and Claude Desktop
- Customize → Connectors → + Add → Add custom connector.
- Name:
Modelfy. URL:https://mcp.modelfy.ai/mcp. - Keep the default sign-in and OAuth client options, and add it. Sign in with your Modelfy account when asked.
- In a conversation, turn it on from + → Connectors.
Custom connectors are available on Free (one connector), Pro, Max, Team and Enterprise. On Team and Enterprise an owner adds it in Organization settings → Connectors, and each member then clicks Connect. Connectors you add on claude.ai also work in Claude Desktop.
ChatGPT
- Open chatgpt.com/plugins → + → Add custom MCP server.
- Name:
Modelfy. Server URL:https://mcp.modelfy.ai/mcp. Authentication: OAuth. - Accept the warning, Create, and install it. Sign in with your Modelfy account.
- In a chat, call it with
@Modelfy.
ChatGPT connects with OAuth only (no API key field), on the web. Custom MCP servers depend on your plan and, in workspaces, on an admin setting; some accounts need Developer mode in Settings.
Cursor
Add to Cursor (one click), or add this to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project):
{
"mcpServers": {
"modelfy": { "url": "https://mcp.modelfy.ai/mcp" }
}
}
Cursor asks you to sign in the first time it uses the server. With an API key, add "headers": { "Authorization": "Bearer ${env:MODELFY_API_KEY}" } and set MODELFY_API_KEY in your environment.
VS Code (GitHub Copilot)
Run MCP: Add Server from the Command Palette, choose HTTP, paste https://mcp.modelfy.ai/mcp, name it modelfy and save it to Copilot Global. Or write ~/.copilot/mcp-config.json:
{
"mcpServers": {
"modelfy": {
"type": "http",
"url": "https://mcp.modelfy.ai/mcp",
"tools": ["*"]
}
}
}
Then click Authenticate in the server's actions and sign in. With an API key, add "headers": { "Authorization": "Bearer ${MODELFY_API_KEY}" } (the braces are required).
Codex (CLI and IDE extension)
codex mcp add modelfy --url https://mcp.modelfy.ai/mcp
Codex starts the sign-in right away. To sign in again later: codex mcp login modelfy. With an API key kept in an environment variable:
codex mcp add modelfy --url https://mcp.modelfy.ai/mcp --bearer-token-env-var MODELFY_API_KEY
Both write ~/.codex/config.toml, which the CLI and the IDE extension share:
[mcp_servers.modelfy]
url = "https://mcp.modelfy.ai/mcp"
# bearer_token_env_var = "MODELFY_API_KEY" # only with an API key
In the IDE extension you can also use the gear → MCP servers → Add server → Streamable HTTP, then Authenticate.
Antigravity (formerly Gemini CLI)
Add this to ~/.gemini/config/mcp_config.json (all workspaces) or .agents/mcp_config.json (one workspace), then run /mcp in the CLI to reload:
{
"mcpServers": {
"modelfy": { "serverUrl": "https://mcp.modelfy.ai/mcp" }
}
}
Sign-in starts on its own the first time. With an API key, add "headers": { "Authorization": "Bearer mfy_live_…" }.
Zed
In your settings, under context_servers:
{
"context_servers": {
"modelfy": { "url": "https://mcp.modelfy.ai/mcp" }
}
}
Without an Authorization header Zed signs you in with OAuth; with an API key, add "headers": { "Authorization": "Bearer mfy_live_…" }.
Cline
In MCP Servers → Remote Servers, or in cline_mcp_settings.json:
{
"mcpServers": {
"modelfy": {
"type": "streamableHttp",
"url": "https://mcp.modelfy.ai/mcp",
"headers": { "Authorization": "Bearer mfy_live_…" }
}
}
}
Your own agent or bot
First get a key: a free sandbox key to build and test, then a live key with device login (nobody copies it) or by hand. Then pick one of these.
Any MCP SDK
TypeScript, with @modelcontextprotocol/client:
import {
Client,
StreamableHTTPClientTransport,
} from "@modelcontextprotocol/client";
const client = new Client({ name: "store-bot", version: "1.0.0" });
await client.connect(
new StreamableHTTPClientTransport(new URL("https://mcp.modelfy.ai/mcp"), {
requestInit: {
headers: { Authorization: `Bearer ${process.env.MODELFY_API_KEY}` },
},
}),
);
const account = await client.callTool({ name: "get_account", arguments: {} });
Read client.getInstructions() once: the server sends its workflow and spending rules there.
Anthropic Messages API (Claude)
The API connects to the server for you. The token goes in authorization_token, without Bearer:
curl https://api.anthropic.com/v1/messages \
-H "content-type: application/json" \
-H "x-api-key: $ANTHROPIC_API_KEY" \
-H "anthropic-version: 2026-10-21" \
-H "anthropic-beta: mcp-client-2025-11-20" \
-d '{
"model": "claude-opus-5-5",
"max_tokens": 1024,
"messages": [{ "role": "user", "content": "What is my Modelfy balance?" }],
"mcp_servers": [{
"type": "url",
"url": "https://mcp.modelfy.ai/mcp",
"name": "modelfy",
"authorization_token": "mfy_live_…"
}],
"tools": [{ "type": "mcp_toolset", "mcp_server_name": "modelfy" }]
}'
OpenAI Responses API
The token goes in authorization, without Bearer, on every request. With a live key its budget limits what the bot can spend; use "require_approval": "always" if a person should approve each call:
{
"model": "gpt-6-astra",
"input": "What is my Modelfy balance?",
"tools": [
{
"type": "mcp",
"server_label": "modelfy",
"server_url": "https://mcp.modelfy.ai/mcp",
"authorization": "mfy_live_…",
"require_approval": "never"
}
]
}
xAI API (Grok)
POST https://api.x.ai/v1/responses:
{
"model": "grok-4.7",
"input": [{ "role": "user", "content": "What is my Modelfy balance?" }],
"tools": [
{
"type": "mcp",
"server_label": "modelfy",
"server_url": "https://mcp.modelfy.ai/mcp",
"headers": { "Authorization": "Bearer mfy_live_…" }
}
]
}
Any other client
If it speaks MCP over Streamable HTTP, give it the URL: it either signs in with OAuth (discovered from the server's 401) or sends the Authorization: Bearer header.