How to Let an Agent Draft Follow-Ups From Reading Sessions
The hardest follow-up email is the one with nothing to say — “just checking in” dressed up five different ways. But if the proposal went out behind a tracked link, there is always something to say: the reading session recorded what they lingered on, what they skipped, and when they came back.
That's a writing brief. With your Apdf workspace connected over MCP, the agent pulls the session and drafts the email — grounded in what the reader actually cared about, phrased so the tracking never shows.
Start from a signal, not a schedule
Follow-ups drafted on day three land wherever day three lands. Better inputs: a filtered automation pinging you on real reads, or the morning triage question surfacing who moved. Today it surfaced Karla: she re-opened the redesign proposal last night.
Let the agent read the session
The session detail the agent worked from — every claim in its reasoning maps to a field:
{
"session": {
"recipient": {
"name": "Karla Weiss",
"email": "karla@studiokarla.example"
},
"created_at": "2026-07-22 21:21:10.000"
},
"metrics": {
"total_duration_ms": 110000,
"pages_viewed": 2,
"total_pages": 5,
"completion_pct": 40,
"downloads": 0,
"prints": 0
},
"pages": [
{ "page": 1, "views": 0, "duration_ms": 0, "viewed": false },
{ "page": 2, "views": 0, "duration_ms": 0, "viewed": false },
{ "page": 3, "views": 1, "duration_ms": 25000, "viewed": true },
{ "page": 4, "views": 1, "duration_ms": 85000, "viewed": true },
{ "page": 5, "views": 0, "duration_ms": 0, "viewed": false }
]
}The draft — grounded, never creepy
Subject: Timeline and payment options for the redesign
Hi Karla,
a thought that might be useful while you're weighing the proposal:
the six-week timeline has some give in how we phase it, and the
investment can follow the phases too — three payments instead of
two works fine on a project this shape.
If it helps to talk any of it through, I have time Thursday.
Best,
JonasNotice what the draft does with the data — and what it refuses to do:
- It answers the pages she re-read — timeline flexibility and payment phasing, because pages 3 and 4 are where her 110 seconds went.
- It never mentions the reading — no “I noticed you opened…”, no timestamps. The tracking decides what to say and when; it must never become the message.
- It stays yours to send — the agent drafts, you approve. A follow-up is a relationship moment; keep a human on the send button.
Scale it without losing the judgment
The loop generalizes: signal in (automation webhook or morning sweep), session detail pulled over MCP, draft out — one per warm reader, each grounded in its own session. Batch it in the morning against the daily digest's list, or wire the CRM-task automation so every qualifying read arrives as a task with the draft attached. The judgment — send, edit, or skip — stays where it belongs.
Where to go from here
The pieces this loop is built from.
Related tutorials
Build a Morning Briefing Agent for Your Documents
A digest lists what happened; a briefing arrives with the thinking done. One standing prompt over the MCP-connected workspace and Claude sweeps every document's sessions on a headless cron — call-first with evidence and a suggested opener, the next tier in one line each, and the silence flagged.
Mint PDF Tracking Links from ChatGPT
The share decision happens mid-conversation, not in a dashboard tab. The MCP server speaks to any client — including ChatGPT — so 'share the pricing guide with Elif' mints her tracked link inside the draft you were writing, and the same chat checks who read and kills a leaver's old link.
Ask Claude Which Deals Are Hot
The morning question isn't about one document — it's 'across everything, who should I call first?' With the workspace connected over MCP, the agent sweeps every document's sessions, ranks re-opens over deep reads over skims, and answers with evidence: the Arslans came back for the price page.