Stacks/Sales

Temporal's Kepler prospecting agent

Kepler is Temporal's in-house prospecting agent: a Claude model picks buying signals from Common Room, Reo.dev, and Exa, a rulebook chooses each email's approach, a model writes the draft from an approved-facts list, and code checks screen it before a rep sends. Day-long scans run as durable Temporal Workflows that survive worker crashes. Claude wrote most of the code.

Jeff LongPublished Nov 9, 2026Last verified Nov 13, 2026
Use case
An internal sales agent that watches buying signals, finds and vets the right contact, and drafts outreach sequences for BDRs and reps to edit and send
Category
Sales
Team size
Built by one GTM operations manager (a former BDR) with help from marketing
Orchestrator / framework
A full scan across every rep's accounts takes most of a day and runs as a Temporal Workflow: each step is an Activity, each signal type a Child Workflow run in a set order, the write step heartbeats every 90 seconds, and 35-minute waits are Timers. Queries report progress and Signals stop a scan cleanly. A rulebook decides the email approach before the model writes, and plain-code rule checks screen every draft. The rep always hits send.
Memory
not stated
Evals
Rule checks in regular code (no AI) block or flag drafts. The author read batches of drafts by hand until the rulebook was in place, and compared drafts against reps' best hand-written emails.
Observability
Temporal Queries show where a scan is.
Cost (stated in writeup)
Not stated
Cost per output (stated)
Not stated in the writeup
WhatWeRun estimate

WhatWeRun estimate

Not enough info to estimate

The post names only 'a Claude model', with no tokens, drafts per scan, or volume.

Reviewed Nov 13, 2026

Observability tools
not stated
Other tools
Common RoomReo.devExaSalesforceOutreachApproved-facts list with sourcesPractice mode and a spend cap
Source
https://temporal.io/blog/ai-prospecting-agent-temporal-kepler

Jeff Long · Published Nov 9, 2026 · Last verified Nov 13, 2026

Related

More like this