Flex Hydra

Code mode

Run Claude Code and Codex in parallel.

Real terminals side by side, one agent each. Flex Hydra tracks which one is working, which one is done and which one is waiting on you, so nothing sits idle unseen.
Four agent panes in a grid, with the workspace rail on the left and the dashboard on the right

Status

Know which agent needs you.

Each pane reports its state. Red needs you: a question, an approval or a crash. Amber is working, green is finished and not yet looked at, gray is idle.

The Dashboard lists the ones waiting on you first, across every window. One click jumps there, and a notification pings you when you’re in another app.

fix upload retryWorking
Claude CodeRead(src/api/upload.ts)Update(src/stores/upload.ts)+ await withRetry(upload)⎿ 2 files changed✻ Running tests…
HUD scoreboardComplete
Claude CodeRead(Source/UI/Scoreboard.h)Edit(Source/UI/Scoreboard.cpp)+ Players.Sort(ByKills);⎿ 1 file changedThe scoreboard sorts by kills now.
e2e checkout testsNeeds you
CodexRan playwright test✓ 12 passed (24.3s)Allow command?$ gh pr create --fill› 1. Yes, proceed
docs cleanupIdle
Grok BuildEdit(docs/setup.md)Edit(README.md)Updated 3 files.› ▍

Flex Hydra · now

e2e checkout tests

Needs your input

The grid

The real CLIs, side by side.

Every pane runs the agent’s own command-line tool, the same one you’d start in a terminal, in a grid you split and resize. Your hooks, skills and MCP servers work as they always do, and quitting an agent leaves a working shell.

Mix them freely: Claude Code on the backend, Codex on the tests, a shell for the dev server.

  • Claude Code
  • Codex
  • Grok Build
  • Gemini CLI
  • Cursor Agent
  • any shell

Workspaces

A workspace is a project’s folders, one agent each, laid out as a grid, side by side or stacked. Open it and the whole setup starts.

Worktrees for Claude Code

Start a Claude Code pane in its own git worktree, so two agents can change the same repo without sharing a checkout.

Pick up where you left off

Reopen a workspace and every pane relaunches and resumes its conversation (Claude Code, Codex, Grok Build).

Recent chats

Past Claude Code and Codex chats for each folder, read from the agents’ own transcripts, including chats started outside Flex Hydra.

One palette for everything

Ctrl/Cmd+Shift+P finds any pane, chat, workspace or action.

Show the agent what it built

Dock a browser beside the grid for your dev server, then screenshot it into the pane you pick.

Give each pane its own account, and keep going at a limitRather chat than type? The same agents in Thread mode

Questions

Running agents in parallel.

Something else? Email help@flexhydra.dev.

Is it a real terminal or a wrapper?

A real terminal. Each pane runs the agent’s own command-line tool in a shell on your machine, so it behaves as it does in your usual terminal, with your settings, hooks, skills and MCP servers.

Which agents can I run?

Claude Code, Codex, Grok Build, Gemini CLI and Cursor Agent, or any shell. Flex Hydra finds the ones installed on your computer.

Can two agents work on the same repo?

Yes. Give each one its own folder in the workspace (a worktree or another checkout), or turn on Worktree when you start a Claude Code pane and it runs in a git worktree of its own.

How do I know which agent needs me?

Each pane’s status shows in its header and on the Dashboard: red needs you, amber is working, green is finished and not yet looked at, gray is idle. A notification pings you when one needs you while you’re in another app.

Do I need API keys?

No. The agents run on your own logins, and their usage counts against the plans you already have. Flex Hydra doesn’t sell model access.

Pricing · one plan

Many heads. One will.

$37/ month

$370/ year, save 17%

Up to 3 computers. Windows and macOS.

Get Flex Hydra Pricing and FAQ