Chained thought input — queue follow-up prompts while a task is in flight
Preflight Checklist
- [x] I have searched existing requests and this feature hasn't been requested yet
- [x] This is a single feature request (not multiple features)
Problem Statement
Right now if Claude Code is mid-task (running tests, executing a command, waiting on a process), any message I type interrupts the current task. That forces a choice: wait silently until it finishes, or break the flow.
What I want is the ability to queue 2–4 follow-up thoughts while Claude is working, so it picks them up in sequence as each task resolves — without me having to watch and wait before I can type the next thing.
The analogy: a surgeon tells the scrub nurse the next three things they'll need before they need them. Nobody waits. Nobody freelances. The sequence is preserved, the work is parallel, and the human stays in flow.
This is the happy medium between:
- Agentic systems — fast, parallel, but they drift from the human's intent and go off on their own
- Single-turn systems — precise and controllable, but bottlenecked by the human having to wait before queuing the next thought
Chained thoughts would preserve the narrative control that makes Claude Code useful, while removing the human bottleneck that breaks creative and technical flow. The creator lays down a chain of intent. Claude follows it in order, resolving each
step, surfacing results, and moving to the next.
That is the thing that makes the difference.
Proposed Solution
Allow chained thoughts. Not chained commands, but chained thoughts that actually use the power of AI to process decisions on outcomes of the previous outcome but insert the chained thought as as guideline. Agentic is powerful because it enables multiple actions but it sucks at retaining vision. No matter what you do you will not be able to cut out humans because it's humans you're making this for. You must find a way to serve the human vision while harnessing the AI's power of thinking and speed to keep it on track.
Alternative Solutions
I don't think there is abother solution. The two need to work together.
Priority
Medium - Would be very helpful
Feature Category
Other
Use Case Example
go see me conversations with your AI at: fe8c3e42-12b6-4f0a-8369-1c1a35c97188
Additional Context
_No response_
This issue has 6 comments on GitHub. Read the full discussion on GitHub ↗