Forkbench
Forkbench is a native macOS terminal and work-management app for running AI coding agents in parallel. It organizes sessions into isolated Threads with durable boards, notebooks and credential controls so people can coordinate agents while keeping code and local terminal data on their Macs.
At a Glance
- Software engineers running several AI coding agents
- Individuals operating agents across more than one Mac
- Developers who want human-in-the-loop review and Git control
- Teams collaborating on software work through shared Thread boards
- +1 more
AI Tools by Forkbench
(1)Forkbench
AI Agent Terminal for macOS
Discussions
No discussions yet
Be the first to start a discussion about Forkbench
Latest News
Signing out leaves nothing behind (v1.7.25): local account data and keys are removed from the Mac on sign-out, with update-installation fixes.
A mark that means what it says (v1.7.24): corrected the terminal-header people mark to distinguish personal and shared Threads.
The Git panel does the whole job (v1.7.23): added merge, rebase, stash and other Git operations, task activity watching and sync fixes.
The activity dot reads as a dial (v1.7.22): introduced richer activity indication for agent thinking, output and handed-off commands.
Products & Services
Native macOS terminal app (macOS 13+) with real login-shell tabs and splits for running Claude Code, Codex, Gemini CLI, Aider, Cursor, Antigravity CLI and custom scripts.
A Thread groups a folder, terminal tabs, notes, secrets and a durable board with one objective and a backlog. Agents can claim, work and hand off tasks, while a human accepts completed work.
Markdown briefs, conventions and decisions organized in notebooks, granted per Thread with read or read-write permissions and synchronized across the user's Macs.
Local-first encrypted storage for API keys, tokens and passwords. Agents can run commands using granted credentials without the credential value entering their prompt, command line or transcript; destination binding can restrict where a key is used.
Market Position
Forkbench positions itself as the coordination, security and persistence layer around terminal-based coding agents rather than as a replacement coding agent or model provider. Its official FAQ contrasts it with Ghostty and iTerm2: those terminals emphasize rendering speed or configurability, while Forkbench emphasizes durable work plans, isolated Threads/worktrees, cross-agent coordination, human acceptance and credentials that agents can use without seeing. It also distinguishes itself from Claude Code Agent Teams by supporting Claude Code, Codex, Gemini CLI, Aider, Cursor, Antigravity CLI and custom scripts side by side.
Founding Story
Forkbench says conventional terminals were designed for one person typing at one prompt, but parallel AI coding agents create tab overload, lost work and coordination problems. It was started to make each piece of agent work a first-class Thread with its own folder, board, notes and secrets, while keeping the human in control of review and acceptance.
Business Model
Revenue Model
Freemium subscription software. The free tier provides the terminal, Threads, worktrees, Notebooks and a one-agent board; Pro subscriptions add Vault, multi-agent boards, Talk, collaboration and multi-Mac sync. Team provides managed per-seat billing, and Enterprise adds SSO, invoicing and procurement terms.
Pricing Tiers
Unlimited Threads, panes and splits; Git branch/worktree management; unlimited synced Notebooks; one-agent board on every Thread; invited-board access; local-first/offline operation.
30-day free trial; Vault, more than one agent on a board, Talk, collaboration, note history and the same board on every Mac. Covers up to 10 of the user's Macs.
Managed per-seat billing for at least two Pro licenses on one invoice; per-seat price can drop as the team grows.
Team functionality plus SSO, invoicing and an organization-specific agreement.
Target Markets
- Software engineers running several AI coding agents
- Individuals operating agents across more than one Mac
- Developers who want human-in-the-loop review and Git control
- Teams collaborating on software work through shared Thread boards
- Developers and organizations concerned about agent access to API keys, tokens and deployment credentials
- Running multiple AI coding agents in parallel without losing track of sessions
- Splitting a software objective into tasks that agents can claim and hand off
- Keeping agents isolated with separate Git worktrees
- Giving agents push, deploy or paid-API access without exposing credential values
- Sharing a Thread board with colleagues while keeping machines, repositories and shell history separate
- Working across multiple Macs from one shared board