Jd Follow Up Email Check

JD Follow-Up Email Check

Purpose

Scan recent Gmail messages, extract job-application outcomes by reading full message bodies (not just subjects), group them by status, log results to a Notion page, and archive rejected records in the Job Applications DB.

Required Inputs

If the user provides different values, use those instead.

Execution Steps

Step 1 — Search Gmail

Use the Gmail MCP search_emails tool with multiple queries in parallel:

Query A — Rejections:

newer_than:{lookback} ("not selected" OR "regret to inform" OR "other candidates" OR "position has been filled" OR "decided to pursue other" OR "will not be moving forward" OR "unsuccessful" OR "unable to offer" OR "no longer under consideration")

Query B — Interview / Moving Forward:

newer_than:{lookback} ("invite you to interview" OR "would like to schedule" OR "phone screen" OR "video interview" OR "next round" OR "advance to" OR "proceed to" OR "we are pleased to" OR shortlisted OR "technical assessment")

Query C — Acknowledgments (application received / under review):

newer_than:{lookback} (subject:"thank you for applying" OR subject:"received your application" OR subject:"thank you for your application" OR subject:"thanks for applying" OR subject:"Welcome to")

Set max: 50 for each query.

Step 2 — Read Full Bodies

For every email returned in Step 1, call get_message to retrieve the full body. Parallel calls are fine — batch up to ~10 at a time.

Step 3 — Classify & Extract

For each message, read the body text (not just subject) and extract:

Classification rules (from body content):

If the body is truncated and ambiguous, classify as Acknowledged and note (body truncated).

Build CompTitle — this is the primary lookup key against the Job Applications DB.

The DB already stores a CompTitle property on every record in the format:

company name|job title

All lowercase, single pipe | separator, no extra spaces around the pipe. Examples from the DB: mastercard|lead product manager- technical, sanofi|product owner, digital portfolio

From each email, build the same format:

1. Extract the company name from the body/sender (use the short brand name, e.g. "Mastercard" not "MasterCard People Services").
2. Extract the job title exactly as it appears in the body (preserve punctuation like hyphens, commas).
3. Lowercase everything, join with |.

Result: "mastercard|lead product manager- technical"

If the job title cannot be extracted, use "company|(unknown role)".

Match quality matters — the CompTitle you build must match the DB value closely enough for the search to find it. Prefer the exact role wording from the email body over a paraphrased version.

Step 4 — Update Notion Follow-Up Email Log

Use notion-fetch to get the current content of the log page (notion_log_page).

Then use notion-update-page with replace_content to write the grouped results.

Content format:

# Follow-Up Email Check — {YYYY-MM-DD}

## Rejected

| Company | Job Title | CompTitle | Email Date |
|---------|-----------|-----------|------------|
| Manulife | Operations Risk & Control Governance Lead | manulife|operations risk & control governance lead | 2026-04-06 |

## Interview Scheduled

| Company | Job Title | CompTitle | Email Date |
|---------|-----------|-----------|------------|
(or "None found in the last {lookback}" if empty)

## Acknowledged / Under Review

| Company | Job Title | CompTitle | Email Date |
|---------|-----------|-----------|------------|
| Amazon | Sr. Manager, Product Manager, Sourcing Performance, SCOT | amazon|sr. manager, product manager, sourcing performance, scot | 2026-04-07 |

Use replace_content so each run produces a clean snapshot (not appended duplicates).

Step 5 — Archive Rejected Records in Job Applications DB

For each rejected application, use the CompTitle built in Step 3 to find and update the matching DB record:

1. Search by CompTitle: Use notion-search with data_source_url: {job_db_data_source} and the email-derived CompTitle as the query (e.g. "sanofi|product owner, digital portfolio").
2. Verify the match: From the search results, fetch the candidate page(s) and check that the record's CompTitle property matches the email-derived CompTitle. This avoids false positives from semantic search returning loosely related records.
3. If a matching record is found:
   • Use notion-update-page → update_properties to set:
     • "Archive": "__YES__"
     • "Status": "Rejected"
   • Record the page URL for the output report.
4. If multiple records share the same CompTitle but have different Job IDs, only update the record whose Job ID matches the email (if the email contains one). If no Job ID is in the email, update all matching CompTitle records.
5. If no match is found, note it as (no matching DB record) in the output.

Step 6 — Output Summary

Print a grouped report to the user:

## Rejected (archived)

| Company | Job Title | DB Record | Archive Updated |
|---------|-----------|-----------|-----------------|
| Manulife | Operations Risk & Control Governance Lead | notion.so/... | Yes |
| Sanofi | Product Owner, Digital Portfolio | notion.so/... | Yes |

## Interview Scheduled

| Company | Job Title |
|---------|-----------|
(none or listed)

## Acknowledged / Under Review

| Company | Job Title |
|---------|-----------|
| Amazon | Sr. Manager, Product Manager, ... |

---
Notion log updated: https://www.notion.so/33c8d41a5cff80d3b905c9b53c6f65e9

Key Rules

1. Always read full message bodies — never classify from subject alone.
2. CompTitle format — "company|job title" all lowercase; this is the lookup key.
3. Only archive on clear rejection — "under review" or "high volume" auto-replies are NOT rejections.
4. Idempotent — running twice on the same day overwrites the Notion log (replace, not append).
5. Parallel where safe — search_emails queries and get_message calls may run in parallel; Notion writes must be sequential.
6. Dedup across queries — the same email may appear in multiple Gmail searches; deduplicate by message ID before classifying.

MCP Tools Used