Forge

Mission Control
for AI-Assisted
Software Development.

Your team runs Claude Code, Cursor, and other AI coding tools — each agent working in parallel, each taking actions in your codebase. Forge is the persistent, visual cockpit that sits on top: seeing everything, enforcing your workflow gates, and letting you supervise or intervene from any device.

Looking for the 2024 desktop app?
Forge is not another AI coding tool. Claude Code and Cursor write code. Forge sits on top of those tools — it's the cockpit, not the agent. It gives your team the situational awareness and control plane that those tools don't provide on their own.
FORGE — CONTROL PLANE Agent Dashboard PERSISTENT · VISUAL · SHARED agent-01 agent-02 Workflow Gates ENFORCE SPEC · TEST · APPROVE ⏸ awaiting approval Audit Trail EVERY ACTION · FULL HISTORY ✓ 47 actions logged Remote Access ANY DEVICE 📱 YOUR AI CODING TOOLS Claude Code writing auth module 63% Cursor refactoring data layer gate: awaiting Claude Code writing tests 87% — green + any tool your choice of agent Forge-agnostic FORGE SUPERVISES WITHOUT REPLACING — YOUR AGENTS KEEP THEIR OWN CONFIG AND CONTEXT
What Forge does

Three capabilities.
One glass cockpit.

Running multiple AI agents across multiple projects without Forge is like managing air traffic with a terminal window. Forge gives your team a shared, persistent view of everything in motion — without changing how the agents themselves work.

01 — VISIBILITY

Visualize concurrent agent work

See every agent, every active task, and every track in a persistent dashboard. Know exactly what each AI session is doing, what phase it's in, and what decisions it's waiting on — without digging through terminal output.

02 — CONTROL

Enforce workflow and quality gates

Your team's spec-first, test-driven process doesn't break when agents are running. Forge surfaces approval prompts and phase checkpoints as native UI elements — agents can't skip gates that matter to you.

03 — AUDIT

Audit and remote supervision

Every action an agent takes is logged and visible. Review diffs, approve changes, and intervene from any device — including mobile. The full conversation and tool-use history stays accessible long after the session ends.

Why it matters

AI agents are fast.
Unsupervised agents are risky.

The productivity gains from AI coding agents are real. The failure modes are also real. Without supervision, agents bypass review, merge code that breaks contracts, and diverge from the process your team spent years building. Forge closes that gap without slowing the agents down.

01.Without Forge

Each engineer monitors their own agent session in a terminal. No shared visibility. Quality gates rely on agent discipline. Diffs get approved without context. Something breaks at 2 a.m. and no one knows which agent or which change caused it.

02.With Forge

One dashboard shows every session, every agent, every phase. Gates are enforced at the platform level. Diffs are reviewable in context, from any device. When something goes wrong, the audit trail tells you exactly what happened and when.

03.Your agents, unchanged

Forge doesn't replace Claude Code or Cursor. It doesn't intercept their calls or modify their output. It observes, gates, and surfaces — then gets out of the way. Your agents keep running the way they already run.

04.Your process, enforced

If your team writes specs before code and tests before merging, Forge makes that a hard gate — not a suggestion. The discipline you've built doesn't evaporate when you add AI speed. It gets enforced at the right moment, automatically.

Get started

Forge is in private early access.

We're working with a small set of engineering teams to validate the supervision model before broader availability. If your team is running AI coding agents at scale, we want to talk.

Looking for the original 2024 Forge desktop app? It's at /forge-classic.