Claude Code opened to any LLM — OpenAI, Gemini, DeepSeek, Ollama, and 200+ models via OpenAI-compatible API shim
TypeScript
docs: rewrite README for OpenClaude — setup instructions for all providers
did:key:z6MkqDnb7Siv3Cwj7pGJq4T5EsUisECqR8KpnDLwcaZq5TPrfd108243ebd5
diff --git a/README.md b/README.mdindex 1853e3e..9a4108e 100644--- a/README.md+++ b/README.md@@ -1,280 +1,222 @@-# Claude Code Source Snapshot for Security Research+# OpenClaude-> This repository mirrors a **publicly exposed Claude Code source snapshot** that became accessible on **March 31, 2026** through a source map exposure in the npm distribution. It is maintained for **educational, defensive security research, and software supply-chain analysis**.+Use Claude Code with **any LLM** — not just Claude.++OpenClaude is a fork of the [Claude Code source leak](https://gitlawb.com/node/repos/z6MkgKkb/instructkr-claude-code) (exposed via npm source maps on March 31, 2026). We added an OpenAI-compatible provider shim so you can plug in GPT-4o, DeepSeek, Gemini, Llama, Mistral, or any model that speaks the OpenAI chat completions API.++All of Claude Code's tools work — bash, file read/write/edit, grep, glob, agents, tasks, MCP — just powered by whatever model you choose.----## Research Context+## Quick Start-This repository is maintained by a **university student** studying:+### 1. Set 3 environment variables-- software supply-chain exposure and build artifact leaks-- secure software engineering practices-- agentic developer tooling architecture-- defensive analysis of real-world CLI systems+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=sk-your-key-here+export OPENAI_MODEL=gpt-4o+```-This archive is intended to support:+### 2. Run Claude Code-- educational study-- security research practice-- architecture review-- discussion of packaging and release-process failures+```bash+claude+```-It does **not** claim ownership of the original code, and it should not be interpreted as an official Anthropic repository.+That's it. The tool system, streaming, file editing, multi-step reasoning — everything works through the model you picked.----## How the Public Snapshot Became Accessible+## Provider Examples-[Chaofan Shou (@Fried_rice)](https://x.com/Fried_rice) publicly noted that Claude Code source material was reachable through a `.map` file exposed in the npm package:+### OpenAI-> **"Claude code source code has been leaked via a map file in their npm registry!"**->-> — [@Fried_rice, March 31, 2026](https://x.com/Fried_rice/status/2038894956459290963)+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=sk-...+export OPENAI_MODEL=gpt-4o+```-The published source map referenced unobfuscated TypeScript sources hosted in Anthropic's R2 storage bucket, which made the `src/` snapshot publicly downloadable.+### DeepSeek----+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=sk-...+export OPENAI_BASE_URL=https://api.deepseek.com/v1+export OPENAI_MODEL=deepseek-chat+```-## Repository Scope+### Google Gemini (via OpenRouter)-Claude Code is Anthropic's CLI for interacting with Claude from the terminal to perform software engineering tasks such as editing files, running commands, searching codebases, and coordinating workflows.+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=sk-or-...+export OPENAI_BASE_URL=https://openrouter.ai/api/v1+export OPENAI_MODEL=google/gemini-2.0-flash+```-This repository contains a mirrored `src/` snapshot for research and analysis.+### Ollama (local, free)-- **Public exposure identified on**: 2026-03-31-- **Language**: TypeScript-- **Runtime**: Bun-- **Terminal UI**: React + [Ink](https://github.com/vadimdemedes/ink)-- **Scale**: ~1,900 files, 512,000+ lines of code+```bash+ollama pull llama3.3:70b----+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_BASE_URL=http://localhost:11434/v1+export OPENAI_MODEL=llama3.3:70b+# no API key needed for local models+```++### LM Studio (local)-## Directory Structure--```text-src/-├── main.tsx # Entrypoint orchestration (Commander.js-based CLI path)-├── commands.ts # Command registry-├── tools.ts # Tool registry-├── Tool.ts # Tool type definitions-├── QueryEngine.ts # LLM query engine-├── context.ts # System/user context collection-├── cost-tracker.ts # Token cost tracking-│-├── commands/ # Slash command implementations (~50)-├── tools/ # Agent tool implementations (~40)-├── components/ # Ink UI components (~140)-├── hooks/ # React hooks-├── services/ # External service integrations-├── screens/ # Full-screen UIs (Doctor, REPL, Resume)-├── types/ # TypeScript type definitions-├── utils/ # Utility functions-│-├── bridge/ # IDE and remote-control bridge-├── coordinator/ # Multi-agent coordinator-├── plugins/ # Plugin system-├── skills/ # Skill system-├── keybindings/ # Keybinding configuration-├── vim/ # Vim mode-├── voice/ # Voice input-├── remote/ # Remote sessions-├── server/ # Server mode-├── memdir/ # Persistent memory directory-├── tasks/ # Task management-├── state/ # State management-├── migrations/ # Config migrations-├── schemas/ # Config schemas (Zod)-├── entrypoints/ # Initialization logic-├── ink/ # Ink renderer wrapper-├── buddy/ # Companion sprite-├── native-ts/ # Native TypeScript utilities-├── outputStyles/ # Output styling-├── query/ # Query pipeline-└── upstreamproxy/ # Proxy configuration+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_BASE_URL=http://localhost:1234/v1+export OPENAI_MODEL=your-model-name```----+### Together AI-## Architecture Summary--### 1. Tool System (`src/tools/`)--Every tool Claude Code can invoke is implemented as a self-contained module. Each tool defines its input schema, permission model, and execution logic.--| Tool | Description |-|---|---|-| `BashTool` | Shell command execution |-| `FileReadTool` | File reading (images, PDFs, notebooks) |-| `FileWriteTool` | File creation / overwrite |-| `FileEditTool` | Partial file modification (string replacement) |-| `GlobTool` | File pattern matching search |-| `GrepTool` | ripgrep-based content search |-| `WebFetchTool` | Fetch URL content |-| `WebSearchTool` | Web search |-| `AgentTool` | Sub-agent spawning |-| `SkillTool` | Skill execution |-| `MCPTool` | MCP server tool invocation |-| `LSPTool` | Language Server Protocol integration |-| `NotebookEditTool` | Jupyter notebook editing |-| `TaskCreateTool` / `TaskUpdateTool` | Task creation and management |-| `SendMessageTool` | Inter-agent messaging |-| `TeamCreateTool` / `TeamDeleteTool` | Team agent management |-| `EnterPlanModeTool` / `ExitPlanModeTool` | Plan mode toggle |-| `EnterWorktreeTool` / `ExitWorktreeTool` | Git worktree isolation |-| `ToolSearchTool` | Deferred tool discovery |-| `CronCreateTool` | Scheduled trigger creation |-| `RemoteTriggerTool` | Remote trigger |-| `SleepTool` | Proactive mode wait |-| `SyntheticOutputTool` | Structured output generation |--### 2. Command System (`src/commands/`)--User-facing slash commands invoked with `/` prefix.--| Command | Description |-|---|---|-| `/commit` | Create a git commit |-| `/review` | Code review |-| `/compact` | Context compression |-| `/mcp` | MCP server management |-| `/config` | Settings management |-| `/doctor` | Environment diagnostics |-| `/login` / `/logout` | Authentication |-| `/memory` | Persistent memory management |-| `/skills` | Skill management |-| `/tasks` | Task management |-| `/vim` | Vim mode toggle |-| `/diff` | View changes |-| `/cost` | Check usage cost |-| `/theme` | Change theme |-| `/context` | Context visualization |-| `/pr_comments` | View PR comments |-| `/resume` | Restore previous session |-| `/share` | Share session |-| `/desktop` | Desktop app handoff |-| `/mobile` | Mobile app handoff |--### 3. Service Layer (`src/services/`)--| Service | Description |-|---|---|-| `api/` | Anthropic API client, file API, bootstrap |-| `mcp/` | Model Context Protocol server connection and management |-| `oauth/` | OAuth 2.0 authentication flow |-| `lsp/` | Language Server Protocol manager |-| `analytics/` | GrowthBook-based feature flags and analytics |-| `plugins/` | Plugin loader |-| `compact/` | Conversation context compression |-| `policyLimits/` | Organization policy limits |-| `remoteManagedSettings/` | Remote managed settings |-| `extractMemories/` | Automatic memory extraction |-| `tokenEstimation.ts` | Token count estimation |-| `teamMemorySync/` | Team memory synchronization |--### 4. Bridge System (`src/bridge/`)--A bidirectional communication layer connecting IDE extensions (VS Code, JetBrains) with the Claude Code CLI.--- `bridgeMain.ts` — Bridge main loop-- `bridgeMessaging.ts` — Message protocol-- `bridgePermissionCallbacks.ts` — Permission callbacks-- `replBridge.ts` — REPL session bridge-- `jwtUtils.ts` — JWT-based authentication-- `sessionRunner.ts` — Session execution management--### 5. Permission System (`src/hooks/toolPermission/`)--Checks permissions on every tool invocation. Either prompts the user for approval/denial or automatically resolves based on the configured permission mode (`default`, `plan`, `bypassPermissions`, `auto`, etc.).--### 6. Feature Flags--Dead code elimination via Bun's `bun:bundle` feature flags:--```typescript-import { feature } from 'bun:bundle'--// Inactive code is completely stripped at build time-const voiceCommand = feature('VOICE_MODE')- ? require('./commands/voice/index.js').default- : null+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=...+export OPENAI_BASE_URL=https://api.together.xyz/v1+export OPENAI_MODEL=meta-llama/Llama-3.3-70B-Instruct-Turbo```-Notable flags: `PROACTIVE`, `KAIROS`, `BRIDGE_MODE`, `DAEMON`, `VOICE_MODE`, `AGENT_TRIGGERS`, `MONITOR_TOOL`+### Groq----+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=gsk_...+export OPENAI_BASE_URL=https://api.groq.com/openai/v1+export OPENAI_MODEL=llama-3.3-70b-versatile+```-## Key Files in Detail+### Mistral-### `QueryEngine.ts` (~46K lines)+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=...+export OPENAI_BASE_URL=https://api.mistral.ai/v1+export OPENAI_MODEL=mistral-large-latest+```-The core engine for LLM API calls. Handles streaming responses, tool-call loops, thinking mode, retry logic, and token counting.+### Azure OpenAI-### `Tool.ts` (~29K lines)+```bash+export CLAUDE_CODE_USE_OPENAI=1+export OPENAI_API_KEY=your-azure-key+export OPENAI_BASE_URL=https://your-resource.openai.azure.com/openai/deployments/your-deployment/v1+export OPENAI_MODEL=gpt-4o+```-Defines base types and interfaces for all tools — input schemas, permission models, and progress state types.+----### `commands.ts` (~25K lines)+## Environment Variables-Manages registration and execution of all slash commands. Uses conditional imports to load different command sets per environment.+| Variable | Required | Description |+|----------|----------|-------------|+| `CLAUDE_CODE_USE_OPENAI` | Yes | Set to `1` to enable the OpenAI provider |+| `OPENAI_API_KEY` | Yes* | Your API key (*not needed for local models like Ollama) |+| `OPENAI_MODEL` | Yes | Model name (e.g. `gpt-4o`, `deepseek-chat`, `llama3.3:70b`) |+| `OPENAI_BASE_URL` | No | API endpoint (defaults to `https://api.openai.com/v1`) |-### `main.tsx`+You can also use `ANTHROPIC_MODEL` to override the model name. `OPENAI_MODEL` takes priority.-Commander.js-based CLI parser and React/Ink renderer initialization. At startup, it overlaps MDM settings, keychain prefetch, and GrowthBook initialization for faster boot.+---++## What Works++- **All tools**: Bash, FileRead, FileWrite, FileEdit, Glob, Grep, WebFetch, WebSearch, Agent, MCP, LSP, NotebookEdit, Tasks+- **Streaming**: Real-time token streaming+- **Tool calling**: Multi-step tool chains (the model calls tools, gets results, continues)+- **Images**: Base64 and URL images passed to vision models+- **Slash commands**: /commit, /review, /compact, /diff, /doctor, etc.+- **Sub-agents**: AgentTool spawns sub-agents using the same provider+- **Memory**: Persistent memory system++## What's Different++- **No thinking mode**: Anthropic's extended thinking is disabled (OpenAI models use different reasoning)+- **No prompt caching**: Anthropic-specific cache headers are skipped+- **No beta features**: Anthropic-specific beta headers are ignored+- **Token limits**: Defaults to 32K max output — some models may cap lower, which is handled gracefully----## Tech Stack--| Category | Technology |-|---|---|-| Runtime | [Bun](https://bun.sh) |-| Language | TypeScript (strict) |-| Terminal UI | [React](https://react.dev) + [Ink](https://github.com/vadimdemedes/ink) |-| CLI Parsing | [Commander.js](https://github.com/tj/commander.js) (extra-typings) |-| Schema Validation | [Zod v4](https://zod.dev) |-| Code Search | [ripgrep](https://github.com/BurntSushi/ripgrep) |-| Protocols | [MCP SDK](https://modelcontextprotocol.io), LSP |-| API | [Anthropic SDK](https://docs.anthropic.com) |-| Telemetry | OpenTelemetry + gRPC |-| Feature Flags | GrowthBook |-| Auth | OAuth 2.0, JWT, macOS Keychain |+## How It Works++The shim (`src/services/api/openaiShim.ts`) sits between Claude Code and the LLM API:++```+Claude Code Tool System+ |+ v+ Anthropic SDK interface (duck-typed)+ |+ v+ openaiShim.ts <-- translates formats+ |+ v+ OpenAI Chat Completions API+ |+ v+ Any compatible model+```++It translates:+- Anthropic message blocks → OpenAI messages+- Anthropic tool_use/tool_result → OpenAI function calls+- OpenAI SSE streaming → Anthropic stream events+- Anthropic system prompt arrays → OpenAI system messages++The rest of Claude Code doesn't know it's talking to a different model.----## Notable Design Patterns+## Model Quality Notes-### Parallel Prefetch+Not all models are equal at agentic tool use. Here's a rough guide:-Startup time is optimized by prefetching MDM settings, keychain reads, and API preconnect in parallel before heavy module evaluation begins.+| Model | Tool Calling | Code Quality | Speed |+|-------|-------------|-------------|-------|+| GPT-4o | Excellent | Excellent | Fast |+| DeepSeek-V3 | Great | Great | Fast |+| Gemini 2.0 Flash | Great | Good | Very Fast |+| Llama 3.3 70B | Good | Good | Medium |+| Mistral Large | Good | Good | Fast |+| GPT-4o-mini | Good | Good | Very Fast |+| Qwen 2.5 72B | Good | Good | Medium |+| Smaller models (<7B) | Limited | Limited | Very Fast |-```typescript-// main.tsx — fired as side-effects before other imports-startMdmRawRead()-startKeychainPrefetch()-```+For best results, use models with strong function/tool calling support.-### Lazy Loading+----Heavy modules (OpenTelemetry, gRPC, analytics, and some feature-gated subsystems) are deferred via dynamic `import()` until actually needed.+## Files Changed from Original-### Agent Swarms+```+src/services/api/openaiShim.ts — NEW: OpenAI-compatible API shim (724 lines)+src/services/api/client.ts — Routes to shim when CLAUDE_CODE_USE_OPENAI=1+src/utils/model/providers.ts — Added 'openai' provider type+src/utils/model/configs.ts — Added openai model mappings+src/utils/model/model.ts — Respects OPENAI_MODEL for defaults+src/utils/auth.ts — Recognizes OpenAI as valid 3P provider+```-Sub-agents are spawned via `AgentTool`, with `coordinator/` handling multi-agent orchestration. `TeamCreateTool` enables team-level parallel work.+6 files changed. 786 lines added. Zero dependencies added.-### Skill System+----Reusable workflows defined in `skills/` are executed through `SkillTool`. Users can add custom skills.+## Origin-### Plugin Architecture+This is a fork of [instructkr/claude-code](https://gitlawb.com/node/repos/z6MkgKkb/instructkr-claude-code), which mirrored the Claude Code source snapshot that became publicly accessible through an npm source map exposure on March 31, 2026.-Built-in and third-party plugins are loaded through the `plugins/` subsystem.+The original Claude Code source is the property of Anthropic. This repository is not affiliated with or endorsed by Anthropic.----## Research / Ownership Disclaimer+## License-- This repository is an **educational and defensive security research archive** maintained by a university student.-- It exists to study source exposure, packaging failures, and the architecture of modern agentic CLI systems.-- The original Claude Code source remains the property of **Anthropic**.-- This repository is **not affiliated with, endorsed by, or maintained by Anthropic**.+This repository is provided for educational and research purposes. The original source code is subject to Anthropic's terms. The OpenAI shim additions are public domain.