MCP Overview

The Sybilion MCP is an AI-powered time series forecasting connector. You give it your historical monthly data, and it returns forward-looking predictions enriched by external signals. It automatically correlates your series against trillions of economic, financial, and market data points to identify the most relevant drivers. It allows you to interact with the Sybilion API through natural language commands.

This guide describes the MCP capabilities, integrations with AI clients like ChatGPT or Claude, and explain the best practices.

What it does

• Forecasts monthly time series up to 12 months ahead, with confidence intervals (80% and 90% bands)
• Validates your data before submission, so you know upfront whether your series meets quality requirements
• Backtests the model against the last 12 months to assess reliability
• Explains what drove the forecast via feature importance and Granger correlations on external signals (economic indicators, sector news, market data, etc.)
• Visualizes results as ready-to-render charts with history, forecast median, and confidence bands

It is made for analysts, data teams, procurement teams, and business intelligence users who need monthly forecasts without the need of building or maintaining ML infrastructure.

MCP integrations

The Sybilion MCP server exposes the full public API surface as tools that any MCP-compatible AI client can discover and call.

MCP server URL: https://mcp.sybilion.dev/mcp
Authentication: OAuth — the client handles the flow, you just need to approve access in your browser when requested.

Once connected, your AI client can submit forecasts, poll status, retrieve driver recommendations, and browse regions and categories. All through natural-language conversation backed by the actual API.

---

ChatGPT

Requirements

ChatGPT custom connectors require Developer Mode and a Plus, Pro, Business, Enterprise, or Edu plan. Business and above plans require an admin to enable Developer Mode for the organisation first.

Setup

1. Open Settings → Apps → Advanced settings and enable Developer Mode.
2. Go to Settings → Apps → Advanced Settings and click Create App.
3. Fill in the connector:
   • Name: Sybilion
   • Description: Submit forecasts, get driver recommendations, and browse the Sybilion API.
   • Connector URL: https://mcp.sybilion.dev/mcp
   • Authentication: Keep OAuth as default. Also confirm that you want to use the Sybilion MCP by checking the box 'I understand and want to continue'.
4. Click Create. ChatGPT connects to the server and lists the advertised tools — you should see forecast, drivers, regions, categories, and account related tools.
5. Complete the OAuth approval in the browser pop-up that follows.

Using it

In any new chat, click + near the composer → More → select Sybilion. The connector is active for that conversation. Write-tool calls (forecast submit, driver recommend) require per-call approval unless you choose to remember approvals for the session.

Example prompts

• "Using the Sybilion API, submit a 6-month forecast for this monthly data and tell me when the job settles."
• "Use the Sybilion API to recommend drivers for energy commodities in Europe and rank them by importance."
• "Check my Sybilion account balance and tell me if I have enough credit to run another forecast."

---

Claude

Claude.ai and Claude Desktop

Both the web UI and desktop app support remote MCP connectors.

Individual accounts (Pro or Max):

1. Go to Customize → Connectors.
2. Click + → Add custom connector.
3. Enter the server URL: https://mcp.sybilion.dev/mcp
4. Click Add, then complete the OAuth flow that opens in your browser.
5. Enable the connector per conversation via + → Connectors in the chat composer.

Team or Enterprise accounts:

An Owner must first add the connector at Organization settings → Connectors, then individual members connect to it from their own Customize → Connectors page.

Example prompts

Once the connector is active in a conversation:

• "Submit a monthly forecast for this data and poll until it settles."
• "Recommend drivers for Commodities (category 46) in Europe (region 3) with recency factor 0.7."
• "Show me my current balance and tier."

---

TradingView Remix

TradingView Remix: AI Chart Copilot is a Chrome extension with a built-in MCP client. You can add the Sybilion MCP server so it calls the Sybilion API during chart analysis conversations.

Setup

1. Install the TradingView Remix extension from the Chrome Web Store and sign in.
2. Open TradingView, click the Remix icon to open the side panel.
3. In the side panel, open Settings → MCP Servers (or the equivalent MCP integration section).
4. Add a new MCP server:
   • URL: https://mcp.sybilion.dev/mcp
   • Auth: OAuth
5. Approve the OAuth flow in the browser pop-up.
6. Ask the extension to list available Sybilion tools to confirm the connection:

   "List the tools available from the Sybilion MCP server."

Example prompts

• "Using the Sybilion API, which drivers correlate most with this chart's commodity? Filter by the visible region."
• "Use the Sybilion API to submit a 6-month forecast for the series in this chart and show me the result when it settles."

TIP

Sybilion forecasts are async — the MCP tool polls until the job completes, then returns the result. Runs can take tens of seconds to a few minutes.

If the agent cannot retrieve forecast data

Some MCP clients pass tool arguments as a plain string instead of a structured dict. If the agent fails to retrieve a forecast or artifact, ask it to call the tool again with the correct argument structure:

Tool	Required arguments
get_forecast	{ "job_id": "<uuid>" }
get_forecast_chart	{ "job_id": "<uuid>" }
get_forecast_artifact	{ "job_id": "<uuid>", "artifact_name": "<artifact_name>" }

---

See also

• Quickstart — first forecast in five steps.
• Authentication — API keys vs session tokens.
• Public OpenAPI — full spec and schema reference.
• Forecasts · Drivers — feature walkthroughs.