Google Colab MCP

This folder is a small sandbox to understand and test the new Google Colab MCP server, plus a playful notebook project to see how well an agent can drive Colab end-to-end.

What is the "Colab MCP" (Google)?

Google released an open-source Colab MCP (Model Context Protocol) Server. The core idea is: an MCP-compatible AI agent can programmatically control Google Colab (not just run code somewhere else).

Instead of copying snippets from a terminal into Colab cells, the agent can:

• Manage notebook dependencies (for example, installing libraries at runtime)
• Create and edit notebook content (including writing new .ipynb files and markdown cells)
• Write and execute Python cells in the notebook
• Reorganize content so the final notebook reads like a finished workflow

In short: Colab becomes a fast, reproducible, cloud sandbox that the agent can actively build inside.

Repo setup (your local MCP server config)

This repo uses a standard MCP-style config file: colabMcp/.mcp.json

Current config:

{
  "mcpServers": {
    "colab-mcp": {
      "command": "uvx",
      "args": ["git+https://github.com/googlecolab/colab-mcp"],
      "timeout": 30000
    }
  }
}

So, when your MCP-capable client starts the server, it uses uvx to run the googlecolab/colab-mcp package.

Prerequisites (as described by the upstream project/blog):

• uv
• git
• python

Notebook — SEC EDGAR 10-Q Financial Report Analyzer

What it does

It fetches and analyzes SEC EDGAR 10-Q financial data end-to-end by:

1. Searching for a company (by name or ticker) to get its CIK
2. Fetching the most recent 10-Q filing history
3. Pulling structured financial metrics from SEC companyfacts (XBRL)
4. Building tables + multiple trend/risk charts
5. Optionally running a narrative analysis using a local LLM via Ollama

Notebook sections at a glance

You can view the fully rendered notebook (with outputs & charts) directly on nbviewer using the badge above, or expand the preview below.

| # | Section | Description | |---|---------|-------------| | 1 | Company Search | Look up any public company by name or ticker | | 2 | Select Company & Fetch 10-Q Filings | Choose from results and pull filing history | | 3 | Extract Financial Data from Company Facts | Parse XBRL data for key metrics (Revenue, Net Income, Assets, EPS, etc.) | | 4 | Financial Data Table | Tabular summary of quarterly financials | | 5 | Revenue & Profitability Trend Charts | Matplotlib visualizations of revenue, income, and profitability over time | | 6 | Quarter-over-Quarter Growth Analysis | Calculate and chart QoQ growth percentages | | 7 | Key Financial Ratios | Compute ratios like profit margin, asset-to-liability, etc. | | 8 | YoY Growth Comparison Chart | Year-over-year comparison visualizations | | 9 | Filing Links & Raw Data Access | Direct links to SEC filings and raw data | | 10 | Ollama AI Gateway — Setup | Install and configure a local LLM (Ollama) on Colab | | 11 | Ollama AI Gateway — Client | Python client to interact with the Ollama API | | 12 | AI Financial Analysis | LLM-generated narrative analysis of the financial data | | 13 | Parallel Analysis Demo | Batch-mode analysis of multiple companies simultaneously | | 14 | AI-to-Chart Algorithm | Automated radar chart, scorecard, and risk heatmap from AI output | | 15 | Parallel Multi-Company Analysis | Side-by-side comparison dashboard across companies |

# Install dependencies
!pip install requests pandas matplotlib plotly -q

import requests
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker
import json
import time
from datetime import datetime

# SEC EDGAR requires a User-Agent header with contact info
HEADERS = {
    "User-Agent": "SECAnalyzer/1.0 (research@example.com)",
    "Accept": "application/json"
}

BASE_URL = "https://efts.sec.gov/LATEST"
DATA_URL = "https://data.sec.gov"

def search_company(query, max_results=10):
    """Search for a company by name or ticker using SEC EDGAR."""
    url = f"{BASE_URL}/search-index?q={query}&dateRange=custom&startdt=2020-01-01&forms=10-Q"
    resp = requests.get(url, headers=HEADERS)
    tickers_url = "https://www.sec.gov/files/company_tickers.json"
    tickers_resp = requests.get(tickers_url, headers=HEADERS)
    tickers_data = tickers_resp.json()

    results = []
    query_lower = query.lower()
    for key, entry in tickers_data.items():
        name = entry.get("title", "").lower()
        ticker = entry.get("ticker", "").lower()
        if query_lower in name or query_lower == ticker:
            results.append({
                "cik": str(entry["cik_str"]).zfill(10),
                "ticker": entry.get("ticker", "N/A"),
                "name": entry.get("title", "Unknown")
            })
        if len(results) >= max_results:
            break
    return results

# Key financial metrics to extract from 10-Q filings
METRICS = {
    "Revenues": ["Revenues", "RevenueFromContractWithCustomerExcludingAssessedTax",
                  "SalesRevenueNet", "RevenueFromContractWithCustomerIncludingAssessedTax"],
    "NetIncome": ["NetIncomeLoss", "ProfitLoss"],
    "TotalAssets": ["Assets"],
    "TotalLiabilities": ["Liabilities"],
    "StockholdersEquity": ["StockholdersEquity",
                           "StockholdersEquityIncludingPortionAttributableToNoncontrollingInterest"],
    "CashAndEquivalents": ["CashAndCashEquivalentsAtCarryingValue",
                           "CashCashEquivalentsAndShortTermInvestments"],
    "OperatingIncome": ["OperatingIncomeLoss"],
    "EarningsPerShare": ["EarningsPerShareBasic"],
    "GrossProfit": ["GrossProfit"],
}

Results it saves (Colab)

The notebook writes its outputs into:

• /content/{TICKER}_10Q_financial_data.csv
• /content/{ticker}_ai_dashboard.png (per-company AI dashboard)
• /content/comparison_dashboard.png (multi-company comparison dashboard)

How to run it (in Colab)

1. Open Financial_Report_Analyzer.ipynb in Google Colab
2. Ensure the SEC User-Agent header includes your contact info (the notebook includes a placeholder you should replace if needed)
3. Run cells from top to bottom

Because the SEC requires a descriptive User-Agent, leaving any placeholder can lead to blocks or rate limiting.

Packages used (quick list)

• uv + git + python (for running the MCP server)
• Notebook deps:
  • requests
  • pandas
  • matplotlib
  • plotly
  • plus Ollama install/setup steps (inside the notebook)

Why this exists

This is intentionally for fun: it's a practical way to verify the Colab MCP server can drive a non-trivial notebook workflow—install dependencies, execute Python, and produce real artifacts—without manual copy/paste between local tooling and Colab.

All Rights Reserved — TheKingHippopotamus — 2026