clisesh

English | 简体中文

clisesh is a terminal UI designed to help developers quickly find and restore previous agent CLI working environments. If you frequently juggle multiple long-running tasks across different AI agent CLIs (Kiro, Antigravity, Claude Code, Codex, OpenCode, etc.), clisesh aggregates all your active and historical sessions in one place, lets you instantly search through them with AI-generated task summaries, and seamlessly resumes your workflow with a single keystroke.

By fully unlocking the value of historical sessions, clisesh eliminates redundant prompt context setup, boosting development efficiency while significantly reducing token consumption.

Features

• Unified Session Hub: Aggregates sessions across Kiro, Antigravity (agy), Claude Code, OpenAI Codex, and OpenCode in a single, newest-first view.
• Smart AI Summaries: Automatically generates concise task summaries using your preferred local CLI or cloud LLM backend. Press m on any session to manually edit, customize, or override descriptions on the fly.
• One-Key Resume: Select any session and press Enter to seamlessly take over and restore the CLI environment right where you left off.
• Active Session Safety: Automatically detects active running sessions with a lock indicator (●) and protects them against accidental deletion.
• Real-Time Filtering & Batch Operations: Quick search across titles, descriptions, and session IDs with /. Multi-select sessions with Space for batch operations.

Supported Providers

clisesh discovers and aggregates installed agent CLIs automatically (provider = "all").

Provider	ID	Resume Command	Active Protection
Kiro	kiro	kiro-cli chat --resume-id <id>	✅ Auto-detected (●)
Antigravity	antigravity / agy	agy --conversation <id>	✅ Auto-detected (●)
Claude Code	claude	claude --resume <id>	✅ Auto-detected (●)
OpenAI Codex	codex	codex resume <id>	✅ Auto-detected (●)
OpenCode	opencode	opencode --session <id>	✅ Auto-detected (●)

You can launch clisesh for a specific provider using --provider <id> or by setting provider = "..." in the configuration file.

Installation

Download Pre-built Binaries (Recommended)

Download the archive for your operating system and architecture directly from the GitHub Releases page:

• Linux: x86_64 (amd64) or arm64
• macOS: Apple Silicon (arm64) or Intel (amd64)

Extract the archive and move clisesh into your PATH:

tar -xzf clisesh_*.tar.gz
sudo mv clisesh /usr/local/bin/

Build from Source

Requires Go 1.22+.

git clone https://github.com/semigodking/clisesh
cd clisesh
make build          # produces bin/clisesh (static binary)

Quick Start & Usage

clisesh                       # Launch TUI aggregating all providers (default)
clisesh tui --provider kiro   # Launch TUI for a specific provider
clisesh list --provider all   # List sessions (non-interactive, scriptable)
clisesh summarize [id...]     # Pre-generate summaries in background
clisesh config init           # Generate default configuration file (~/.config/clisesh/config.toml)
clisesh version               # Print version

Keyboard Shortcuts

Key	Action	Description
↑ / k, ↓ / j	Navigate	Move cursor up / down
g / home, G / end	Jump	Move to top / bottom of list
pgup / pgdown	Scroll	Page up / down
space	Mark	Toggle multi-selection for batch operations
enter	Resume	Resume selected session in its agent CLI
m	Edit	Manually edit description in a modal dialog
r	Re-summarize	Queue selected session(s) for re-summarization
d	Delete	Delete session (active ● sessions are protected)
i	Info	View session details (ID, age, CWD, provider)
e	Error	View detailed error logs if summarization fails
/	Filter	Search across titles, descriptions, and IDs (esc to cancel)
q / ctrl+c	Quit	Exit clisesh

Configuration

clisesh works out of the box with zero configuration. To customize its behavior, generate a configuration file:

clisesh config init

Configuration is loaded from ~/.config/clisesh/config.toml (or $XDG_CONFIG_HOME/clisesh/config.toml).

Example Configuration

# Active provider: "all" (default), "kiro", "antigravity", "claude", "codex", or "opencode"
provider = "all"

# Color theme: "default", "gray", "high-contrast", or "256"
theme = "default"

[summarizer]
# Subprocess CLI backend (e.g. "subprocess.kiro", "subprocess.agy")
# or HTTP LLM backend (e.g. "llm.openai", "llm.deepseek", "llm.glm")
backend = "subprocess.kiro"
batch_size = 5                   # Concurrent background summarization workers

[summarizer.subprocess.kiro]
command = "kiro-cli"
args = ["chat", "--no-interactive", "--model", "{model}", "--effort", "low", "--wrap", "never", "{prompt}"]

[summarizer.llm.openai]
base_url = "https://api.openai.com/v1"
api_key = "${OPENAI_API_KEY}"
model = "gpt-4o-mini"
rpm = 60                         # Max requests per minute limit

See configs/config.example.toml for all available options, including custom color overrides, token extraction settings, and LLM backends.

Documentation & Development

• Design Overview — Architecture and component design
• Provider Capability Matrix — On-disk layouts and capabilities
• Component Specifications — Detailed specs per module

make test     # Run unit tests
make lint     # Run code formatting and vet checks
make dist     # Cross-compile release archives for Linux & macOS

Donation

If you think this software is helpful, consider to donate.

• 
• paypal.me/redsocks2

License

MIT