Forge: My Own Claude Code
Multi-model, multi-provider, zero vendor lock-in — an agentic coding platform from scratch
TypeScript · Node.js · SQLite · 8 LLM Provider
I rebuilt Claude Code. Not out of boredom, not out of hubris — but because I hit a wall that everyone who seriously works with AI agents knows: vendor lock-in.
The Problem Nobody Talks About
One Tool, One Provider
- Claude Code = Anthropic only
- GPT = OpenAI only
- Gemini = Google only
- Best model per task? Impossible
- 370,000 LOC tied to one vendor
Forge: Any Model, Any Task
- Opus for architecture
- Gemini for 1M token context
- GPT-5 for reasoning
- Ollama for local models
- Model router selects automatically
Forge: A Smithy for Software
An autonomous development platform that self-optimizes, uses any LLM, and is controllable from any device. The architecture: a model router abstracts all providers. Each task gets not the most expensive model, but the most suitable one.
Reverse-Engineered the Source Code
File Tools
Read, Write, Edit, Glob, Grep — fully rebuilt.
Shell Tools
Bash execution with timeout, background, sandbox.
Agent Tools
Spawn sub-agents, track tasks, isolate worktrees.
MCP Integration
Tool discovery, schema resolution, server management.
Workflow Tools
Cron scheduling, hook system, plan mode.
Extras
Cost tracking, LLM summarization, tool blocking via hooks.
Features Claude Code doesn't have: cron scheduling, LLM-based context summarization, cost tracking with budgets per session/day/model, hook-based tool blocking.
Cross-Device: Plan on Phone, Build on Desktop
Discuss an idea on your phone in Claude.AI. Sit at the computer, rebuild context. Session ends, knowledge lost. In Forge, session state lives on the server. Phone for planning. Desktop for deep work. Tablet for monitoring. Context never lost.
Bootstrapping: Forge Builds Forge
“Every improvement to Forge makes Forge better at improving itself.”
Why This Matters for Everyone
Anyone seriously developing with AI today needs access to multiple models, persistence across sessions, quality gates that enforce — and the freedom to choose the best tool for the task.
“The future of AI development isn't one model, one tool, one provider. It's an orchestra.”
— Philipp
Multi-provider AI architecture for your organization? Let's talk about vendor-independent strategies.
Book a consultation