Subscription Audit

Search Gmail and Slack via MCP to discover, extract, and report on all IT/AI/SaaS subscription fees.

Workflow

The audit involves these steps:

• Search Gmail with tiered queries (high-yield first, then broader)

• Search Slack for subscription references

• Read thread details to extract exact amounts

• Consolidate and deduplicate results

• Generate the final audit report

Step 1: Search Gmail

Use gmail_search_messages via MCP. Run queries from references/search-queries.md , starting with Tier 1 (highest yield). Set max_results to 50–100 per query.

manus-mcp-cli tool call gmail_search_messages --server gmail --input '{"q": "<QUERY>", "max_results": 100}'

For each result file, extract thread IDs and subjects with Python:

import json with open('<result_file>') as f: data = json.load(f) for t in data.get('result', {}).get('threads', []): for msg in t.get('messages', []): h = msg.get('pickedHeaders', {}) print(f"{t['id']} | {h.get('subject','')} | {h.get('from','')}")

Key Tier 1 queries (always run these):

• from:stripe.com receipt

• from:github.com receipt payment

• from:googleplay-noreply@google.com subscription

• from:anthropic.com OR from:mail.anthropic.com receipt invoice

• from:openai.com receipt invoice billing

• from:ionos OR from:godaddy OR from:namecheap OR from:cloudflare invoice domain

For the full query list with rationale, read references/search-queries.md .

Step 2: Search Slack

Use slack_search_public_and_private via MCP:

manus-mcp-cli tool call slack_search_public_and_private --server slack --input '{"query": "subscription invoice payment billing receipt", "limit": 20}'

Run 2–3 query variants. Slack rarely contains billing data but may reference costs in discussion.

Step 3: Read Thread Details

For threads identified as IT/AI subscriptions, read full content to extract amounts:

manus-mcp-cli tool call gmail_read_threads --server gmail --input '{"thread_ids": ["<id1>","<id2>"], "include_full_messages": true}'

Batch up to 10 thread IDs per call. From the plain text content, extract:

• Amount — look for £ , $ , € followed by digits

• Plan name — e.g. "Claude Pro", "Starter", "Team Plan"

• Billing period — e.g. "Jan 18 – Feb 18, 2026"

• Payment method — e.g. "Visa ending 8149"

• VAT — typically 20% for UK-based users

Step 4: Consolidate

Optionally run the consolidation script on saved MCP result files:

python3 /home/ubuntu/skills/subscription-audit/scripts/consolidate_results.py <output_dir> <json1> [json2 ...]

Group results by service provider. Deduplicate by thread ID. For each service, record:

Field Source

Service name Sender name or subject line

Category AI Assistant, Developer Tools, Cloud/Hosting, AI Voice, etc.

Current plan From receipt body

Monthly cost From receipt body (convert USD→GBP if needed)

Payment method From receipt body

Billing cycle Monthly or Annual

First seen / last seen Earliest and latest receipt dates

Step 5: Generate Report

Use templates/report_template.md as the structural guide. The report must include:

• Executive Summary — service count, total monthly/annual cost, summary table

• Detailed Analysis — one section per service with billing history

• Conclusion — recommendations on cost optimisation

Look up the current USD→GBP exchange rate when converting amounts. Deliver the report as a Markdown file with the raw JSON data as a secondary attachment.

IT/AI Service Categories

When classifying subscriptions, use these categories:

Category Examples

AI Assistant Anthropic Claude, OpenAI ChatGPT, Perplexity

AI Developer Tools GitHub Copilot, Cursor, Replit

AI Voice / Media ElevenLabs, Midjourney, Runway

Developer Tools GitHub Plans, JetBrains, VS Code

Cloud / Hosting AWS, Azure, Hetzner, DigitalOcean, Vercel

Web Hosting / Domain IONOS, GoDaddy, Namecheap, Cloudflare

AI / Cloud Storage Google AI Pro, iCloud+

Productivity Notion, Figma, Slack Pro

Security / Network Tailscale, VPN services