Skip to content
HomeNewsDigestsConceptsGuidesToolbox
AboutSubscribeUA
Subscribe

AI Today Brief

The daily AI-engineering brief. Built in public. EN · UA.

XTelegramLinkedInYouTubeRSS

Follow AI Today Brief on LinkedIn for daily AI-engineering updates and the weekly “5 shifts that changed how developers work” PDF.

Explore

NewsDigestsConceptsGuides

Company

SubscribeAdvertiseAbout

Legal

Editorial policyAI disclosurePrivacyTerms

© 2026 AI Today Brief. All rights reserved.

  1. Home/
  2. News/
  3. Tools & releases/
  4. Agent-shell 0.78 Adds In-Flight Steering and Persistent Prompt Queuing
Tools & releases

Agent-shell 0.78 Adds In-Flight Steering and Persistent Prompt Queuing

The agent-shell package for Emacs introduces mid-turn prompt steering using the Agent Client Protocol alongside a persistent writable prompt. Developers can course-correct running Claude and Codex agents without canceling current execution.

September 24, 2026· 4 min read
OKCurated by Oleksandr Kuzmenko, AI Product Engineer·Updated September 24, 2026·Sources cited on every story
AI-assisted · editor-reviewed·How we use AI
Agent-shell 0.78 Adds In-Flight Steering and Persistent Prompt Queuing

Impact: Medium

Why it matters

You can steer long-running coding agents mid-execution instead of aborting runs or waiting for bad tool loops to finish.

TL;DR

  • 01Steer Claude and Codex mid-turn using M-RET without killing the session context
  • 02Enable session token cost monitoring using agent-shell-show-cost-indicator
  • 03Drag and drop files and macOS screenshot previews directly into agent-shell buffers

Key facts

Release version
agent-shell 0.78
Protocol extension used
_session/steering (ACP)
Supported steering agents
Claude, Codex

Mid-Turn Steering via Agent Client Protocol Agent-shell 0.78 leverages the _session/steering protocol extension to redirect agent execution without terminating turns. Submitting prompts with M-RET triggers agent-shell-busy-submit-override-function to redirect models mid-flight. Standard RET submissions route through agent-shell-busy-submit-default-function to queue subsequent instructions cleanly. ### Persistent Prompt and Workflow Integrations The new writable prompt remains active at the end of the buffer at all times via agent-shell-persistent-prompt-enabled. Users can drop image thumbnails directly from macOS screenshot utilities into the session buffer, displaying an image preview while attaching other file formats as @path references. ### Session Cost and Tool Enhancements Developers tracking API expenditures can toggle cumulative session costs in buffer headers with (setopt agent-shell-show-cost-indicator t). The release also packages agent-shell-org-config for managing agent skills in org-roam, adds transient permission interfaces, and supports Google Antigravity and Qoder CLI runtimes.

Try it in 2 minutes

(setopt agent-shell-persistent-prompt-enabled t)
(setopt agent-shell-show-cost-indicator t)

elisp

✓ When to use

  • Managing autonomous multi-step coding agents inside Emacs
  • Preventing token waste when an agent chooses a wrong directory or execution strategy

✕ When NOT to use

  • When using IDEs outside the Emacs ecosystem such as Cursor or VS Code
  • When connected to agents that do not implement the Agent Client Protocol

What to do today

  • →Update agent-shell to version 0.78 in your Emacs package manager
  • →Set (setopt agent-shell-show-cost-indicator t) to track session token budgets
  • →Use M-RET instead of RET when you need to redirect an agent handling an incorrect loop
#agent-shell#Emacs#Claude#Codex#Antigravity#Qoder CLI

Sources

  • agent-shell 0.78 updates
ShareShare on XShare on LinkedIn
← Previous storyAgent Orchestration Shifts from "Can Run" to "Can Run to Completion"Next story →Meta Opens Muse Agent Platform to Developer Connectors and Mac Automation

Related stories

  • Tools & releasesAlibaba Releases Qwen Image 2.1: Compact 7B Open-Weight Generator Rivals Frontier Models
  • Tools & releasesAnthropic Orchestrates 950 Claude Code Agents for Autonomous Dataset Discovery
  • Tools & releasesNVIDIA SWE-Serve: About One in Three Agent Patches Fail Live Serving
  • Tools & releasesAgent Orchestration Shifts from "Can Run" to "Can Run to Completion"

Email digest

Get the morning AI brief

One email a day — the stories that matter for engineers, founders and tech leads. Human-edited, with links to primary sources.

  • ✓120+ sources scanned daily
  • ✓Edited by a human
  • ✓1 email per day
  • ✓EN + UA

By subscribing you agree to the privacy policy.