Weekly Reporting (LangGraph + Swytchcode)

A LangGraph agent that turns a Google Sheet of metrics into a published Notion page and an emailed weekly report.

Run one command to pull your weekly numbers, publish them to Notion, and email stakeholders, without writing API glue code or managing credentials and retries.

What this does

This demo automates a recurring weekly report. It reads a range of metrics from a Google Sheet, creates a Notion page summarizing them, and emails an HTML table of the same numbers to a stakeholder address through Resend. The three steps run as a LangGraph state machine, so each node passes its result to the next.

Every external call goes through Swytchcode, a deterministic API execution layer for AI agents. The agent code never calls Google, Notion, or Resend directly. Instead it asks the Swytchcode runtime to run a named method, and the runtime validates the request against a schema registry of 2,000+ integrations, handles auth and retries, and records an audit trail of what ran.

How it works

The graph has three nodes and runs them in order:

load_report -> create_notion_page -> email_stakeholders

• load_report reads the configured range from your Google Sheet via spreadsheets.values:batchget.get and stores the rows in state.
• create_notion_page writes a new page into your Notion database via pages.page.create, with the metrics rendered as a paragraph block.
• email_stakeholders sends an HTML table of the metrics to the report address via emails.email.create (Resend), including a reference to the Notion page that was created.

Prerequisites

• Python 3.9+

• Swytchcode CLI: install with the verified script for your platform:

  npm install -g swytchcode

• A Google Sheets access token, a Notion integration token, and a Resend API key (see the table below).

Setup

1. Clone the repo:

   git clone https://github.com/swytchcodehq/Weekly-Reporting-Langgraph.git
   cd Weekly-Reporting-Langgraph

2. Install dependencies:

   pip install -r requirements.txt

3. Copy the example env file and fill in your keys:

   cp .env.example .env

4. Fetch the integrations declared in .swytchcode/tooling.json:

   swytchcode bootstrap

Environment variables

Variable	Required	Description
GOOGLE_ACCESS_TOKEN	Yes	OAuth bearer token the Sheets read is authenticated with.
GOOGLE_SPREADSHEET_ID	Yes	Spreadsheet ID, taken from the sheet URL.
GOOGLE_SHEETS_RANGE	No	A1 range to read. Defaults to Sheet1!A1:B10.
NOTION_API_KEY	Yes	Notion integration token (secret_...).
NOTION_DATABASE_ID	Yes	ID of the Notion database the report page is created in.
RESEND_API_KEY	Yes	Resend API key (re_...).
REPORT_EMAIL	Yes	Address that receives the emailed report.
SWYTCHCODE_TOKEN	Yes	Swytchcode auth token, from the Swytchcode dashboard under Settings, API keys.

The first row of the sheet range is treated as a header and skipped; remaining rows are read as metric, value pairs.

Run

python main.py

Expected output

The script prints each node as it runs and a summary at the end:

[1/3] Loading report data from Google Sheets for Week of June 10, 2026...
    Loaded 5 metrics
[2/3] Creating Notion report page for Week of June 10, 2026...
    Notion page created: 1a2b3c4d-...
[3/3] Emailing report to you@example.com...
    Email sent to you@example.com

Weekly report complete!
   Rows reported:   5
   Notion page ID:  1a2b3c4d-...
   Email sent:      True

After a run you should see a new page in your Notion database, titled for the week you ran it, and a report email in the REPORT_EMAIL inbox.

Canonical IDs used

Service	Canonical ID
Google Sheets	spreadsheets.values:batchget.get
Notion	pages.page.create
Resend	emails.email.create

Part of the Swytchcode demo collection

Runnable LangGraph + Swytchcode examples:

• Customer-Onboarding-Langgraph
• Create-And-Send-Payment-Langgraph
• Lead-Qualification-Langgraph
• Bug-Escalation-Langgraph

License

MIT. See LICENSE.