FollowAgents AI AGENTS DIRECTORY
Home / All Agents / Agent Session Manager
Dev & Engineering terminal-session-managementtmuxmulti-agent-workflowsconversation-resumegit-diff-viewinghistory-searchbubble-teagolang

Agent Session Manager

Organize, monitor, and resume multiple AI coding CLI sessions from one terminal interface.

View on GitHub ↗ Quick start ↓
Source repo
izll/agent-session-manager
Stars
★ 16
Last updated
1mo ago
License
MIT
Primary language
Go
FA score
64/100 · Some gaps

At a glance

How it runs
CLI
Works with
Universal · cross-platformCodex · Claude Code
Cost
Free software; you pay for model usage
Setup effort
Medium · a few setup steps
You'll need
tmux on Linux/macOS or psmux on WindowsAt least one supported AI CLI tool or a custom CLI commandShell / CLINetwork accessLocal filesystem
Typical use
A developer running Claude Code, Codex, and Gemini CLI concurrently can monitor each task's busy, waiting, idle, or stopped state from one TUI.
Not a fit if
  • Users unwilling to install tmux or psmux, which core session execution depends on
  • Users who want a graphical UI or mobile push notifications; use ASMGR Desktop instead
  • Users expecting the terminal and desktop editions to share one session list; their stores are separate
Source review
64/100 · Some gaps

What does this agent do, and when should you use it?

Agent Session Manager (ASMGR) is a local terminal UI built with Go and Bubble Tea that controls AI coding assistants through tmux, or psmux on Windows. It directly supports Claude Code, Gemini CLI, Aider, OpenAI Codex, Amazon Q, and OpenCode, while also accepting custom CLI commands. Sessions can be isolated by project, arranged into groups, run in parallel from the same directory, and expanded with multiple agent or shell tabs. The interface provides ANSI-aware live output, per-tab activity detection, prompt delivery, conversation resumption, history search, and Git diff inspection. Its deployment boundary is the user's terminal, local repositories, filesystem configuration, and terminal multiplexer; it is not a hosted agent service.

Running asmgr opens a project selector and loads persisted state from projects.json, the default sessions.json, or project-specific projects/<project>/sessions.json files under ~/.config/agent-session-manager/. Components under session/ handle process lifecycles, PTY and tmux integration, storage, project records, status detection, and discovery of agent conversation histories. The ui/ package uses Bubble Tea, Bubbles, and Lip Gloss to render the session list, previews, overlays, split view, help, and color picker. After the user chooses Claude, Gemini, Aider, Codex, Amazon Q, OpenCode, or a custom command, ASMGR launches and controls that process in a tmux/psmux session or tab, captures its output for live preview, and can send prompts to it without attaching. It discovers resumable conversations for Claude, Gemini, Codex, OpenCode, and Amazon Q, while global search reads supported Claude, Aider, OpenCode, and shell history stores. In Git working directories, it produces either a Session diff measured from HEAD at session start or a Full diff containing all uncommitted changes, and it persists session paths, agent types, resume IDs, colors, groups, notes, and auto-yes settings.

  1. A developer running Claude Code, Codex, and Gemini CLI concurrently can monitor each task's busy, waiting, idle, or stopped state from one TUI.
  2. An engineer handling several tasks in one repository can launch named parallel instances of the same agent without merging their conversation contexts.
  3. A solo developer maintaining many repositories can separate sessions by project and organize long-lived work with collapsible groups, favorites, and notes.
  4. A developer reviewing agent output can switch from the live preview to Session or Full Git diffs without leaving the manager.
  5. A Claude Code user exploring risky or competing implementations can fork one conversation into a new tab or a separate session.
  6. A terminal-heavy user returning to old work can search supported agent and shell histories and resume compatible conversations.

How do you install or deploy this agent?

The runtime requires tmux on Linux/macOS or the tmux-compatible psmux on Windows. Install at least one supported AI CLI, unless the session will run a custom command. On macOS or Linux, install with Homebrew:

brew tap izll/tap
brew install asmgr

Linux users can instead run the quick installer:

curl -fsSL https://raw.githubusercontent.com/izll/agent-session-manager/main/install.sh | bash

Building from source requires Go 1.24 or newer:

git clone https://github.com/izll/agent-session-manager.git
cd agent-session-manager
go build -o asmgr .
cp asmgr ~/.local/bin/

On Windows, download asmgr_<version>_windows_amd64_setup.exe from Releases. The per-user installer adds asmgr to PATH and can offer to install psmux where winget is available; open a new terminal after installation.

How do you use this agent?

First ensure tmux or psmux and the chosen AI CLI are available on PATH, then start the application:

asmgr

Select or create a project, or continue without one. In the session list, press n to create a session, Enter to start and attach, or s to start without attaching; a offers replacement or a named parallel instance. Use t to add an agent or ordinary terminal tab, p to send a prompt to a running process, and r to browse resumable conversations for supported agents. Press D for the diff view, F to choose Session or Full diff, and Ctrl+F for global history search. Once attached, Ctrl+q detaches quickly; F1 or ? opens the complete in-application help.

What are this agent's strengths and limitations?

Pros
  • One interface directly manages six named AI coding CLIs plus arbitrary custom commands, avoiding a single-provider session model.
  • Parallel instances in the same directory, multi-tab sessions, project isolation, groups, favorites, and persistent notes support large ongoing workloads.
  • ANSI-aware live previews, per-tab activity states, prompt sending, and conversation resumption centralize observation and control.
  • Session and Full Git diffs, split-screen comparison, and cross-history search add review and retrieval workflows around agent execution.
  • Distribution options include Homebrew, a Windows installer, a Linux script, deb and rpm packages, and source builds.
Limitations
  • Core session execution depends on tmux; Windows users need psmux or must point ASMGR_TMUX at tmux under WSL or Cygwin.
  • Each AI CLI must still be installed and configured separately, including its credentials, network access, and provider charges; ASMGR supplies no model service.
  • Feature coverage differs by agent: Aider and custom commands cannot resume conversations, and conversation forking is Claude-only.
  • Global history search covers Claude, Aider, OpenCode, and terminal history only, and it is restricted to ASMGR project directories.
  • The terminal and desktop editions keep separate session stores, so adopting both does not provide a shared session list.

How does this agent compare with similar options?

The project explicitly says it was inspired by Claude Squad. Unlike a session manager centered on one assistant, ASMGR documents direct support for Claude Code, Gemini CLI, Aider, OpenAI Codex, Amazon Q, OpenCode, and custom commands. The separate ASMGR Desktop repository is the named graphical alternative: it uses Wails and Svelte and adds real terminals, a diff viewer, a file browser, and mobile push notifications, but its sessions are stored separately from the TUI edition.

Key facts side by side with the most closely related agents.

Agent Source review Form / cost Stars Updated Language Full support on
Agent Session Manager This agent 64 · Some gaps CLIFree + model costs ★ 16 1mo ago Go Codex · Claude Code
Agent Deck 74 · Some gaps CLIFree + model costs ★ 1.2k today Go Codex · Claude Code
Agent Manager 75 · Good CLIFree ★ 646 1d ago Go Codex · Claude Code
Agent of Empires 79 · Good CLIFree ★ 3.9k today Rust Codex · Claude Code

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Some gaps
64/ 100 5-point scale 3.2 / 5
Trust 11/29
Reliability 8/14
Adaptability 15/18
Convention 14/18
Effectiveness 12/13
Verifiability 4/8
Why each dimension lost points
Trust11 / 29 · 1.9/5

The code and README show that the app manages tmux/psmux sessions, reads agent histories, can send prompts, and offers updates and YOLO mode. Updates and pattern refreshes use the network, and YOLO can reduce agent approval constraints. The README describes configuration and some data sources, and updates require confirmation, but permission scope, protection of sensitive histories, dependency auditing, external effects, and recovery are not covered thoroughly, so scores are limited. The release workflow and MIT license are visible, but publisher identity and maintenance responsibility are not established.

Reliability8 / 14 · 2.9/5

The README and main.go explain the multiplexer prerequisite and provide understandable startup and update errors; go.mod pins direct and indirect dependency versions. The README documents features and shortcuts, but some key descriptions overlap or conflict (for example, Ctrl+F for history search and Ctrl+f for tab tracking). Static evidence cannot establish ongoing dependency availability.

Adaptability15 / 18 · 4.2/5

The documentation clearly targets terminal developers using multiple AI coding CLIs and covers Linux, macOS, Windows, tmux, psmux, and custom commands. It states some capability boundaries, such as Aider lacking resume support and forking being Claude-only; code also describes when pattern refresh occurs. Limitations across all agents and environments are not fully documented, so this is not full marks.

Convention14 / 18 · 3.9/5

The README is organized around installation, operation, configuration, and features, with platform-specific installation steps, shortcut tables, configuration examples, and some limitations. The license file clearly provides MIT terms. The documentation shows v0.7.5, but the supplied materials do not establish a versioning policy, changelog, or clear maintenance responsibility, which limits those scores.

Effectiveness12 / 13 · 4.6/5

The feature set addresses organization, preview, resume, search, and diff review for multi-agent terminal sessions. The README provides usable shortcuts and installation commands. The benefit is clear, but the product requires a separate multiplexer and agent CLIs, while self-updating, history search, and YOLO add operational and data-management costs, so cost-benefit is not full marks.

Verifiability4 / 8 · 2.5/5

README installation and configuration claims can be compared with main.go's startup check, network pattern refresh, and update confirmation; go.mod and the release workflow provide limited corroboration. Many feature claims are supported only by the README, and the packet does not include corresponding implementation or tests. The documentation generally distinguishes supported agents and limitations, but not enough to verify every behavior.

Risks and how to mitigate them
  • YOLO mode may change an agent's approval or sandbox behavior; check the effect for the specific agent before enabling it.
  • History search reads local agent conversation histories and shell histories, which may contain sensitive information.
  • The README offers a curl-piped installer and built-in self-update; the supplied evidence does not show download verification or update rollback details.
Evidence confidence: Low Reviewed Nov 08, 2026 Reviewed revision e8a3be8edfcd
Review evidence README.mdmain.goLICENSEgo.modworkflows/release.yml
See the full review method →
Maintain this project? Add the FARS badge to your README

The badge shows the current review score and verdict and links back to this page. It refreshes about once an hour after a re-review.

FARS

[![FARS review](https://followagents.com/badge/agent-session-manager.svg)](https://followagents.com/en/agents/agent-session-manager)
<a href="https://followagents.com/en/agents/agent-session-manager"><img src="https://followagents.com/badge/agent-session-manager.svg" alt="FARS review"></a>

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

  1. v0.9.2
    Release notes excerpt
    See [CHANGELOG.md](https://github.com/izll/agent-session-manager/blob/main/CHANGELOG.md) for what changed in this release, described from a user's point of view. Changelog * 198fadb6502194a7f8c45d34155d1188ae10f5e3 docs(changelog): describe the 0.9.2 release * 8c5ec4a2d957bf232bd2f84cacfee0be903d1dc6 feat(codex): start Codex without its background server * 77a21e5cbb6b1370f14fa1cf87ceb914d42d60ec fix(codex): use the bypass flag Codex still has Full Changelog: https://github.com/izll/agent-session-manager/compare/v0.9.1...v0.9.2
    View on GitHub ↗
  2. v0.9.1
    Release notes excerpt
    See [CHANGELOG.md](https://github.com/izll/agent-session-manager/blob/main/CHANGELOG.md) for what changed in this release, described from a user's point of view. Changelog * 80fb0be2c370c2d6b0430e54bedf2082c101d9c9 docs(changelog): describe the 0.9.1 release * 533181b6246d960795662dad6dcd3b63e76a3062 fix(status): show Codex's work instead of its model name * 7829be4b8ae72ce2ca890704db8842adf82228a7 fix(tmux): set window options with -w so a dead pane stays
    View on GitHub ↗
  3. v0.9.0
    Release notes excerpt
    See [CHANGELOG.md](https://github.com/izll/agent-session-manager/blob/main/CHANGELOG.md) for what changed in this release, described from a user's point of view. Changelog * f0140b6eab14d8e37f422cb1e3d42512ff85c51c build: add a Windows installer that offers to install psmux * 624f166d62eea74024acd6e58ee6e98af4083b74 feat: run on Windows, through psmux
    View on GitHub ↗

FAQ

Does ASMGR include or host an AI model?
No. It launches already installed AI CLIs or custom commands. Accounts, authentication, usage charges, and network requirements remain the responsibility of those underlying tools.
Can it run on Windows?
Yes. The Windows installer uses native psmux and does not require WSL or MSYS2. Alternatively, ASMGR_TMUX can point to an absolute tmux path under WSL or Cygwin.
Which agents support conversation resume?
Claude, Gemini, Codex, OpenCode, and Amazon Q are documented as resumable. Aider and custom commands are not.
Are sessions isolated between projects?
Yes. Each project has its own session list and groups. A per-project single-instance lock also prevents two ASMGR processes from opening the same project simultaneously.
Does ASMGR modify source code itself?
The launched AI CLI or custom command performs any code changes. ASMGR manages its process, terminal, input, preview, and persisted metadata, and can display session-scoped or complete uncommitted Git changes.
Was this page helpful?
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Open Source Coding Agents, Compared

Open source coding agents ranked by FARS review — score, trust, license, language and maintenance side by side, with what each one is and isn't good at.

Claude Code Agents, Compared

Claude Code agents ranked by FARS review — compatibility is read from each agent's own AGENTS.md and documentation, not assumed, and only full support is listed.