aattaran/deepclaude

Use Claude Code's autonomous agent loop with DeepSeek V4 Pro, OpenRouter, or any Anthropic-compatible backend. Same UX, 17x cheaper.

What it solves

DeepClaude allows users to use the autonomous agent loop of Claude Code while replacing the expensive Anthropic models with more affordable alternatives like DeepSeek V4 Pro, OpenRouter, or Fireworks AI. This significantly reduces the cost of running an autonomous coding agent without sacrificing the core functionality of the tool loop, file editing, and bash execution.

How it works

The project provides a local proxy that intercepts API calls from the Claude Code CLI. By manipulating environment variables (such as ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN), it redirects model requests to a different backend. The proxy can switch backends mid-session via a control endpoint, allowing users to switch between DeepSeek and Anthropic (e.g., for more complex reasoning tasks) without restarting the terminal.

Who it’s for

Developers who use Claude Code and want to reduce their API costs or use different model providers while maintaining the same user experience and and autonomous agent capabilities.

Highlights

  • Multi-backend support: Supports DeepSeek, OpenRouter, Fireworks AI, and original Anthropic models.
  • Cost tracking: Includes a proxy that tracks token usage and calculates savings compared to Anthropic pricing.
  • Llive switching: Ability to switch backends mid-session using slash commands, CLI flags, or VS Code keyboard shortcuts.
  • Remote control: Enables the use of DeepSeek as the brain for Claude Code sessions opened in a browser.
  • Tool compatibility: Maintains support for file reading/writing, bash execution, git operations, and subagent spawning.

Related

  • Project
    askalf/dario

    A local proxy that routes AI tool requests through a Claude Pro/Max subscription to avoid per-token API billing.

  • Project
    raine/claude-code-proxy

    A proxy that allows the Claude Code coding harness to use alternative LLM providers like OpenAI Codex, Kimi, Grok, and Cursor Agent by translating Anthropic API traffic.

  • Project
    routatic/proxy

    A Go-based CLI proxy that routes Claude Code requests to multiple LLM providers with automatic model selection and format transformation.

  • Project
    OnlyTerp/UltraCode-Shim

    A loopback proxy that enables Claude Code's UltraCode mode to work with any LLM backend, featuring smart routing and orchestrator/worker splitting.

  • Project
    MadAppGang/claudish

    Claudish is a CLI that lets you run Claude Code with any LLM—cloud or local—by proxying requests through a local Anthropic‑compatible server. It supports 580+ OpenRouter models, Gemini, OpenAI, Ollama, and many others, using a simple `provider@model` syntax, and works both online and fully offline.