Using slim.to with OpenAI (ChatGPT & Agents SDK)

Two ways in: the MCP server (Agents SDK / ChatGPT connectors) or the REST API (custom GPT Actions, scripts).

Prerequisite for both: a personal API key.

Approach 1 — MCP server

OpenAI Agents SDK (Python)

The Agents SDK runs local stdio MCP servers directly, so your agent can upload files it just wrote:

from agents import Agent, Runner
from agents.mcp import MCPServerStdio

async def main():
    async with MCPServerStdio(
        params={
            "command": "uvx",
            "args": ["--from", "/path/to/slimto/mcp", "slimto-mcp"],
            "env": {"SLIMTO_API_KEY": "slim_your_key_here"},
        }
    ) as slimto:
        agent = Agent(
            name="Analyst",
            instructions=(
                "When you produce a report or file the user will share, "
                "upload it with create_link_from_file and give them the short URL."
            ),
            mcp_servers=[slimto],
        )
        result = await Runner.run(agent, "Write the Q3 summary PDF and share it.")
        print(result.final_output)

ChatGPT (web) — hosted connector

slim.to ships a hosted remote MCP server with OAuth:

1. In ChatGPT: Settings → Connectors (enable Developer mode if you don't see the option) → Create / Add connector.
2. Enter the MCP server URL https://slim.to/mcp with OAuth authentication.
3. ChatGPT opens the slim.to sign-in/consent page — click Authorize.

ChatGPT can then shorten URLs, publish HTML/text content it generates as tracked links (create_link_from_content), list your links, and read your analytics. Revoke access anytime from Dashboard → Settings → Connected apps.

The same URL works with the Responses API mcp tool for API-side agents:

tools=[{"type": "mcp", "server_url": "https://slim.to/mcp",
        "authorization": "<oauth-access-token-or-slim_personal-key>"}]

Approach 2 — REST API

Custom GPT Action (works in ChatGPT today)

1. In ChatGPT: Explore GPTs → Create a GPT → Configure → Create new action.
2. Import the schema from https://slim.to/openapi/openapi.json, or paste a minimal spec covering POST /api/v1/links, GET /api/v1/links, and GET /api/v1/analytics/links/{link_id}.
3. Under Authentication choose API Key, header name X-API-Key, and paste your slim_… key.
4. In the GPT's instructions, add:

When the user asks to shorten or share a URL, call createLink and reply with the short_url. When they ask who opened a link, call getLinkAnalytics.

Note: GPT Actions can send JSON, so URL shortening and analytics work great; ChatGPT cannot attach a locally generated binary file to an Action call, so file uploads still need the API from a machine (below) or the web app.

Scripts / anything with HTTP

import requests, os

def slimto_upload(path: str, title: str | None = None) -> str:
    with open(path, "rb") as fh:
        r = requests.post(
            "https://slim.to/api/v1/files",
            headers={"X-API-Key": os.environ["SLIMTO_API_KEY"]},
            files={"file": fh},
            data={"title": title or ""},
        )
    r.raise_for_status()
    return r.json()["link"]["short_url"]

Wire that up as a normal function tool in any OpenAI tool-calling loop; the full endpoint reference is in the API quickstart.

Which approach?