Every run, logged
Agent log
What each agent was asked, what it returned, and what a person did with it. This is the record the learning loop will read.
No ANTHROPIC_API_KEY set: agents run in offline mode, with simple drafts built from the record. Add the key to switch on Claude (claude-sonnet-5-5 for writing, claude-haiku-4-5-20251001 for checks).
Runs logged
0
Emails sent as written
–
0 drafts decided
Tokens
0
0 served from cache
Errors
0
The agents
Host email
Writes host emails from the record on the stage templates. Drafts only.
0 runs
Brand Check
Reviews every draft against the Brand page.
0 runs
Fit suggestion
Suggests the six scores with reasons. A founder decides.
0 runs
Monday priorities
Three priorities per person from to-dos, deadlines and the pipeline.
0 runs
Visual check
Looks at every image next to the photo rules and the reference visuals.
0 runs
Captions
Two or three Instagram captions from Marina's brief and the record.
0 runs
Call notes
Reads the first-call notes and proposes record fields.
0 runs
Reddit reply
Answers the thread first, always discloses, links only where allowed.
0 runs
Substack
Prepares the Substack version of a journal article.
0 runs
Article
Interview transcript to article, SEO fields and fact-check list.
0 runs
Recent runs
No runs yet
Draft an email from a property record to see the first one.