Jeeves: TUI for Browsing and Resuming AI Agent Sessions

What Jeeves Does
Jeeves is a terminal-based tool for managing AI agent conversation history. It aggregates sessions from multiple AI agent frameworks into one interface, allowing developers to browse, search, and resume previous conversations without leaving the terminal.
Key Features
- Browse all AI agent sessions in one place with support for Claude Code, Codex, and OpenCode
- Search sessions by content using plain text or regex patterns
- Preview conversation contents in a split pane view
- Open conversations to read full context
- Resume sessions directly in the agent
Usage
Basic commands:
# Browse all sessions
jeeves
Search for sessions containing "ui refactor"
jeeves ui refactor
Search with a regex pattern
jeeves " fix.*bug "
Keyboard Controls
enter- open sessionr- resume session in agent/- filter sessionsj/k- navigate up/downw- toggle line wrapesc- back to listqorctrl+c- quit
Installation Options
# Homebrew
brew install robinovitch61/tap/jeeves
Nix (with NUR)
nix-shell -p nur.repos.robinovitch61.jeeves
Nix flakes
nix run github:robinovitch61/nur-packages#jeeves
Arch Linux (AUR)
yay -S jeeves-bin
Go install
go install github.com/robinovitch61/jeeves@latest
Windows (winget)
winget install robinovitch61.jeeves
Windows (scoop)
scoop bucket add robinovitch61 https://github.com/robinovitch61/scoop-bucket
scoop install jeeves
Windows (chocolatey)
choco install jeeves
Configuration Tip
By default, Claude Code deletes sessions older than 30 days. To keep all sessions browsable in Jeeves, set a large cleanup period in ~/.claude/settings.json:
{
"cleanupPeriodDays": 99999
}
Technical Details
- Written in Go (99.6% Go, 0.4% Makefile)
- Built with tools from Charm
- Relies on robinovitch61 viewport bubble
- MIT licensed
- Current release: v0.3.0 (April 15, 2026)
Development
To build from source:
git clone [email protected]:robinovitch61/jeeves.git
cd jeeves
go build
📖 Read the full source: HN AI Agents
👀 See Also

Crime Team: Multi-Agent Orchestrator for OpenClaw — Parallel Code Review with Coder Agent
Crime Team v0.1 runs multiple specialist OpenClaw agents in parallel for code review, then integrates findings. Includes per-agent models, a coder agent that applies changes, and a re-audit loop. CLI + GUI.

Manual-Driven Development: A Method to Prevent Claude Code's Confident Divergence
Manual-Driven Development (MDD) is a method that addresses confident divergence in Claude Code, where the AI produces wrong code that passes its own tests. In a production audit, MDD found 190 issues, wrote 876 new tests in under 8 hours, and eliminated rule violations.

Claude-voice: Local TTS with Word Highlighting for Claude Code
Claude-voice is a Python tool that adds local text-to-speech with real-time word highlighting to Claude Code's voice mode. It uses Kokoro TTS (82M parameters) running fully locally without API keys.

Voxlert: Voice Notifications for Claude Code Sessions with Character Voices
Voxlert is a tool that hooks into Claude Code events and speaks notifications using distinct character voices like StarCraft Adjutant, SHODAN, GLaDOS, and HEV Suit. It uses an LLM via OpenRouter to generate in-character lines and runs locally with npm installation.