---
title: "229 Prompts Before I Built the Session Host"
canonical: https://dxdev.com/blog/2026-09-26_session-host-vendor-exit/
datePublished: 2026-09-26
---
I spent 229 prompts asking an editor extension to manage windows before I built a session host instead.

The session ran 20h 16m and opened with a complaint, which the work log truncates mid-word: I liked the Claude extension in Cursor, but when working with Claude it couldn't spawn and manage windows ("...manage these windo…").

## The cost of the extension route

The session title in the log is "Claude extension window management". It ran from 2:25 PM to 10:40 AM, with 229 prompts, out=311,750 tokens and cache_read=69,682,562.

The close-out summary in the log records the decision: the extension route was dropped and our own session host was built instead.

## What the session host does

Every spawned session now gets its own controllable window, with live progress, approvals and restart recovery.

The cockpit can run sessions on the host behind a flag, and one test session has switched over.

The secretary no longer sends refused permission requests to my phone.

## Two tiny sessions in the host project

The host project shows up in the log at 12:37 AM and 12:39 AM. Both sessions have the same first prompt: "Reply with exactly the digit 0." The first took 10 prompts and 3,352 output tokens, the second 9 prompts and 2,619.

## What isn't proven

The follow-ups are filed, and the log marks two things as not yet proven live:

1. The secretary hook on the phone path.
2. Placement ladder steps 2 to 4.

The narrow claim the log supports is this: every spawned session gets its own controllable window with live progress, approvals and restart recovery, and one test session has switched over behind the flag.

## A count worth keeping

The number worth keeping from this session is the prompt count. It opened with a complaint about what a tool couldn't do, and by the close it had taken 229 prompts, ending with the decision to build the thing instead.