[FEATURE] Per-Message Inline Modifiers (model, thinking)

Resolved 💬 4 comments Opened Jan 20, 2026 by whatnickcodes Closed Mar 10, 2026

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

Switching models or toggling thinking mid-conversation requires multiple steps and is painful and is not the future modern:

  • press a keyboard shortcut (Option+P or Option+T)
  • make the selected change, then type your message.

slow and painful

Proposed Solution

Support inline message prefixes that apply model or thinking settings to a single message only, without changing the session default.

Example syntaxes (pick one convention):

# Hashtag style
#haiku what's 2+2?
#opus analyze this architecture
#think work through this carefully
#no-think just format this

# Colon style
haiku: what's the time in tokyo?
opus: review this PR for security issues
think: why is this test flaking?
no-think: convert this to typescript

# Slash inline style
/haiku what dependencies does this use?
/opus design a caching strategy
/think debug this race condition
/no-think rename this variable

# Flag style
--haiku quick syntax check
--opus deep code review
--think step through this algorithm
--no-think summarize in one line

# Bracket style
[haiku] what's this error mean?
[opus] refactor this module
[think] find the bug
[no-think] list the files

# HTML style
[haiku] what's this error mean? [/haiku]
[opus] refactor this module [/opus]

The message uses the specified model/mode, then returns to session defaults for the next message.

Alternative Solutions

_No response_

Priority

Critical - Blocking my work

Feature Category

CLI commands and flags

Use Case Example

_No response_

Additional Context

_No response_

View original on GitHub ↗

This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗