---
title: "Roadmap: what is coming next"
description: "The agent-manager roadmap: big features being built now, next and later for the tmux manager that runs Claude Code, Codex and other AI coding agents."
canonical: "https://agent-manager.dev/roadmap/"
last-updated: "2026-11-01"
section: "Site"
language: "en"
source: "agent-manager"
---

# Roadmap: what is coming next

agent-manager is a free, open-source terminal multiplexer for AI coding agents. It runs Claude Code, Codex, Gemini CLI and 9 more coding CLIs side by side in one tmux terminal UI. This roadmap lists the big features being built for it now, the ones coming next, and the ones planned for later.

Smaller fixes and features live in the [issue tracker](https://github.com/YoanWai/agent-manager/issues). Last updated 1 November 2026.

## Being built now

### Built-in browser pane

A real Chromium browser in a pane next to your agents. You watch the page while an agent drives it through the agent-manager MCP server, and only the agent that opened a browser can drive it.

In progress in [#952](https://github.com/YoanWai/agent-manager/pull/952).

## Coming next

### Plugins

Add your own tools and features to the manager as plugins.

### Panes

Split the manager into panes and work with several agent sessions side by side.

## Planned for later

### Shared setup across agent CLIs

Set up skills, MCP servers, named agents and project memory once, and every agent CLI in the manager uses them, whether a session runs Claude Code, Codex or Gemini CLI.

Each part starts as a research spike: [shared skills](https://github.com/YoanWai/agent-manager/issues/546), [shared MCP servers](https://github.com/YoanWai/agent-manager/issues/547), [named agent definitions](https://github.com/YoanWai/agent-manager/issues/500) and [project memory](https://github.com/YoanWai/agent-manager/issues/498).

### Cost tracking

See what each session and each group spends, across every agent CLI. Today the [stats line](https://agent-manager.dev/docs/stats/) shows CPU and RAM for each agent.

## Shipped

### Messages inbox and commit-by-commit review

Messages sit in an inbox grouped by kind, and the one you pick opens beside the list. Review steps through every commit on the branch, and comments stay on their own commit.

Shipped in [v0.46.0](https://github.com/YoanWai/agent-manager/releases/tag/v0.46.0).

### First-run setup and environment profiles

A first-run setup picks your agents, layout and theme and imports past chats. Sessions launch under named environment profiles, and `agent-manager export` and `import` move your setup to another machine.

Shipped in [v0.45.0](https://github.com/YoanWai/agent-manager/releases/tag/v0.45.0).

### Agents on other machines over SSH

Another machine's sessions join the list over SSH. Attach to them, prompt them, or start an agent there from the same list.

Shipped in [v0.43.0](https://github.com/YoanWai/agent-manager/releases/tag/v0.43.0).

### Native Windows

agent-manager runs natively on Windows, on psmux 3.4 or newer and PowerShell 7.3 or newer.

Shipped in [v0.42.0](https://github.com/YoanWai/agent-manager/releases/tag/v0.42.0).

### Model and reasoning pickers

Pick the model and reasoning effort in the New Session form and in quick prompt mode. Every choice is read live from the CLI, so a new model shows up without an agent-manager update.

Shipped in [v0.40.0](https://github.com/YoanWai/agent-manager/releases/tag/v0.40.0).

## Get involved

Ideas and questions about any of these go in [Discussions](https://github.com/YoanWai/agent-manager/discussions), and a feature that is not here yet is a [feature request](https://github.com/YoanWai/agent-manager/issues/new?template=feature_request.yml). The [contributing guide](https://github.com/YoanWai/agent-manager/blob/main/.github/CONTRIBUTING.md) says how to pick up an issue.