autosteer

🥈Silver

Desktop application that enhances Claude Code with multi-workspace management, persistent sessions, and seamless context switching across macOS, Linux, and Windows.

6690Updated 2mo ago

Intermediate30min to implementautomation

Saves ~480 min per use

Quick InstallView Source

git clone https://github.com/notch-ai/autosteer.git

Works with:

Claude

Overview

About This Skill

AutoSteer is a desktop application built with Electron that extends Claude Code with organized multi-workspace management capabilities. It enables developers to maintain isolated project contexts with persistent sessions, allowing you to switch between different projects without losing conversation history or state. The tool tracks token usage per message and worktree, provides a protocol trace viewer for debugging agent behavior, and supports custom slash commands to extend Claude Code functionality. AutoSteer runs natively on macOS, Linux, and Windows via WSL, storing all configuration locally for privacy and control.

How to Use

[{"step":"Install and configure autosteer in Claude Code by running `claude install autosteer` and following the setup prompts. Ensure you have admin permissions for workspace modifications.","tip":"Use `claude list-workspaces` to verify existing workspaces before running autosteer commands. This avoids conflicts with manually managed workspaces."},{"step":"Define automation rules for your workspaces using the template: `autosteer --create-workspace [WORKSPACE_NAME] --projects [PROJECT_PATH] --rules [AUTOMATION_RULES]`. For example, `autosteer --create-workspace dev-frontend --projects /projects/app --rules 'auto-sync=true,idle_timeout=30'`.","tip":"Start with conservative rules (e.g., longer idle timeouts) and adjust based on performance monitoring. Use `autosteer --list-rules` to review current configurations."},{"step":"Monitor performance and adjust settings using `autosteer --monitor` to view real-time metrics (CPU, memory, idle closures). Use `autosteer --update-settings` to tweak thresholds like CPU limits or sync intervals.","tip":"Enable verbose logging with `autosteer --log-level debug` if issues arise. Logs are stored in `~/.claude/autosteer.log` for debugging."},{"step":"Integrate autosteer with your CI/CD pipeline by adding a step in your workflow (e.g., GitHub Actions) to run `autosteer --sync-workspaces` before deployments. This ensures workspaces are pre-optimized.","tip":"For team environments, use `autosteer --export-config [FILE_PATH]` to share workspace configurations across your team. Import with `autosteer --import-config [FILE_PATH]`."}]

Use Cases

Managing multiple concurrent projects with independent Claude Code sessions and file contexts

Resuming long-running conversations across work sessions while preserving full context

Monitoring token usage and API costs across different worktrees and messages

Debugging Claude Code agent behavior using the protocol trace viewer

Setup & Installation

Quick Install

No install command available. Check the GitHub repository for manual installation instructions.

Alternative Install (Git Clone)

git clone https://github.com/notch-ai/autosteer

Requirements

Claude Code or compatible AI agent
Works with: Claude

Quick Start Guide

Install the Skill

Copy the install command above and run it in your terminal.

Open Your AI Agent

Launch Claude Code, Cursor, or your preferred AI coding agent.

Try It Out

Use the prompt template or examples below to test the skill.

Customize

Adapt the skill to your specific use case and workflow.

Usage Examples

Prompt Template

Use the autosteer skill to automate multi-workspace management in Claude Code. Create a [WORKSPACE_NAME] workspace, add [PROJECT_PATH] as a project, and configure [AUTOMATION_RULES] (e.g., auto-sync on file changes, auto-close idle workspaces after [IDLE_MINUTES] minutes). Ensure all workspaces are optimized for [PERFORMANCE_SETTINGS] (e.g., CPU/GPU allocation, memory limits).

Example Output

I’ve successfully configured autosteer for your multi-workspace setup in Claude Code. Here’s what was accomplished:

1. **Workspace Creation & Configuration**:
- Created a new workspace named `dev-frontend` with a project path set to `/Users/alice/projects/ecommerce-dashboard`.
- Applied automation rules to auto-sync files on changes and auto-close idle workspaces after 30 minutes of inactivity.
- Optimized performance settings with a CPU limit of 80% and memory allocation of 4GB to prevent system slowdowns.

2. **Multi-Workspace Sync**:
- Linked three additional workspaces (`dev-backend`, `qa-testing`, and `docs-review`) to the same Claude Code instance.
- Enabled cross-workspace file sharing, allowing seamless navigation between projects without manual switching.
- Set up a shared `.env` file across all workspaces to maintain consistent environment variables.

3. **Performance Monitoring**:
- autosteer logged the following metrics over the last 24 hours:
- Average CPU usage per workspace: 65% (within the 80% limit).
- Memory usage: 3.2GB (below the 4GB allocation).
- Idle workspace closures: 12 instances, all closed without data loss.
- No critical errors were detected in the logs.

4. **Next Steps**:
- To adjust settings, run `autosteer --update-settings` and modify the YAML config file at `/Users/alice/.claude/autosteer.yaml`.
- For troubleshooting, check the logs at `/Users/alice/.claude/autosteer.log` or run `autosteer --diagnostics`.

This setup ensures your workspaces are always optimized for performance and automation, reducing manual overhead by 70% compared to traditional methods.