forge-orchestrator
Orchestrate Claude Code, Codex CLI, and Gemini CLI on shared repos — single Rust binary, zero deps.
Multi-agent inside a single tool works fine. Claude Code runs 20 subagents and they stay aligned because the tool manages that orchestration internally. The problem is multi-tool: Claude Code, Codex CLI, and Gemini CLI on the same repo with no shared state. That's what the orchestrator solves.
It adds file locking, knowledge capture, task planning, drift detection, and multi-tool orchestration to your development workflow. Three tools reading from and writing to a single state directory. An 11-tool MCP server exposes the full orchestration state to any connected AI client.
The orchestrator exists because I ran two AI tools on the same codebase and watched them fail in the same ways human teams fail. Claude refactored a module. Codex updated tests against the pre-refactor interface. Both saved. Tests failed. Neither tool knew the other existed. I'd spent 23 years watching this exact scenario with talented human teams. The solution was always the same: orchestration infrastructure.
Install
curl -fsSL https://forge.nxtg.ai/install.sh | sh
forge init
Single binary. 4.7 MB. No runtime dependencies.
Quick Demo
Four commands from zero to task board:
forge init
forge config brain openai
forge plan --generate
forge status
From here you can:
forge run — execute tasks headlessly (autonomous mode, great for CI)forge dashboard --pty — launch the TUI dashboard with interactive agent panes
Real-World Result: Stargate Mode
From a 94-line specification, the orchestrator coordinated three AI coding tools to build a complete CLI toolkit:
SPEC.md (94 lines) → forge plan --generate → 15 tasks
Claude Code: 6 design tasks → completed
Codex CLI: 6 implementation → completed
Gemini CLI: 3 test/doc tasks → completed
Result: 5,306 lines of working code
10 auto-committed git entries
13 Python files (CLI + tests)
Zero file conflicts
Three companies. Three tools. One orchestrator. One command:
What You Get
| Feature | What It Does |
|---|
| File locking | Exclusive locks prevent tools from editing the same file simultaneously |
| Knowledge flywheel | Captures decisions, patterns, learnings across tools. Auto-classified and searchable. |
| Plan generation | forge plan --generate decomposes specs into dependency-aware task graphs |
| Drift detection | Compares in-progress work against specs, flags divergence early |
| Task board | Dependency-tracked tasks, tool assignment, progress monitoring |
| Multi-tool adapters | Claude Code (MCP stdio) + Codex CLI + Gemini CLI (filesystem) |
| TUI dashboard | forge dashboard: live tool panes, task status, lock state |
| Headless mode | forge run: autonomous execution for CI/CD pipelines |
| MCP server | 11 tools accessible by any connected AI tool via stdio |
Key Commands
forge init
forge plan --generate
forge dashboard
forge dashboard --pty
forge run
forge status
forge verify
forge uat
forge ship
Architecture
The orchestrator is the policy core of Forge. All governance rules, file locks, and orchestration logic live here. The plugin (L1: Vibe Coder) and UI (L3: Ship Lord) are adapter surfaces. They present the orchestrator's state through different interfaces but don't make policy decisions.
┌──────────────────────────────────────┐
│ forge-orchestrator │
│ (Rust, 4.7 MB, 378 tests) │
│ │
│ File locking · Task planning │
│ Knowledge · Governance · MCP │
│ Multi-tool adapters │
│ │
│ Policy enforced here. │
│ Nowhere else. │
└───────────┬──────────────────┬───────┘
│ │
┌────────┴────────┐ ┌────┴────────┐
│ forge-plugin │ │ forge-ui │
│ (L1 Safety) │ │ (L3 MC) │
│ Claude Code │ │ React │
│ adapter │ │ dashboard │
└─────────────────┘ └─────────────┘
Communication: .forge/ filesystem + MCP stdio. No daemon. No database. State is files.
Tool adapters: Claude Code uses MCP stdio. Codex CLI and Gemini CLI use filesystem conventions. Each tool reads its own config format. No Forge-specific configuration language.
How File Locking Works
File locking exists because I've watched teams lose days to conflicting edits.
When Claude Code starts editing a file, the orchestrator acquires an exclusive lock in .forge/locks/. Codex CLI requesting the same file is queued with a notification of who holds the lock. Locks include timeouts (crashed tools don't hold locks forever) and deadlock detection.
378 tests cover concurrent access, timeout behavior, multi-adapter locking, and edge cases.
How the Knowledge Flywheel Works
The knowledge flywheel exists because I've watched decisions evaporate between sprints.
Every decision, pattern, and learning from tool sessions is captured in .forge/knowledge/. Entries are auto-classified by category. Search across sessions and across tools. Knowledge captured during a Claude Code session is available to Codex CLI the next day.
After a week, the system knows your conventions better than you remember them. New sessions start with context instead of from scratch.
Upgrade Path
The TUI dashboard shows everything. Forge UI shows it better.
Visual dashboards. Governance HUD. Real-time agent collaboration network. And the Infinity Terminal: sessions that survive browser close, network drops, and server restarts. If you've lost a 3-hour agent session to a disconnected SSH pipe, you know why this matters.
git clone https://github.com/nxtg-ai/forge-ui && npm install && npm run dev
58 components. 4,165 tests. 87% coverage.
Links
License
Source available under the Functional Source License 1.1 (FSL-1.1-ALv2). Converts to Apache License 2.0 on 2028-03-18. See LICENSE.md for full terms.