Claude Code's autonomous agent loop (file editing, bash, git, subagent spawning) is the best in class. The problem: it costs $200/month with caps. DeepClaude is a thin shell script that swaps the backend model while keeping the entire tool loop unchanged. Default backend is DeepSeek V4 Pro ($0.44/M input, $0.87/M output) vs Anthropic's $3/$15 per million tokens — a 17x price difference.

How it works

Claude Code reads environment variables for API endpoint and auth. DeepClaude sets these per-session:

• ANTHROPIC_BASE_URL — API endpoint
• ANTHROPIC_AUTH_TOKEN — API key for the backend
• ANTHROPIC_DEFAULT_OPUS_MODEL — model name for Opus-tier tasks
• ANTHROPIC_DEFAULT_SONNET_MODEL — model name for Sonnet-tier tasks
• ANTHROPIC_DEFAULT_HAIKU_MODEL — model name for Haiku-tier (subagents)
• CLAUDE_CODE_SUBAGENT_MODEL — model for spawned subagents

After launch, original settings are restored on exit.

Quick start (2 minutes)

1. Get a DeepSeek API key — sign up at platform.deepseek.com, add $5 credit.

2. Set environment variable:

# Windows (PowerShell)
setx DEEPSEEK_API_KEY "sk-your-key-here"

macOS/Linux
echo 'export DEEPSEEK_API_KEY="sk-your-key-here"' >> ~/.bashrc
source ~/.bashrc

3. Install the script:

# Windows — copy to a directory in PATH
Copy-Item deepclaude.ps1 "$env:USERPROFILE\.local\bin\deepclaude.ps1"

macOS/Linux
chmod +x deepclaude.sh
sudo ln -s "$(pwd)/deepclaude.sh" /usr/local/bin/deepclaude

4. Use it:

deepclaude                              # Launch with DeepSeek V4 Pro (default)
deepclaude --status                     # Show available backends and keys
deepclaude --backend or                 # Use OpenRouter ($0.44/M input)
deepclaude --backend fw                 # Use Fireworks AI (fastest, US servers)
deepclaude --backend anthropic          # Normal Claude Code (for Opus)
deepclaude --cost                       # Show pricing comparison
deepclaude --benchmark                  # Latency test across all providers

Supported backends

Backend	Flag	Input/M	Output/M	Servers	Notes
DeepSeek (default)	--backend ds	$0.44	$0.87	China	Auto context caching (120x cheaper on repeat turns)
OpenRouter	--backend or	$0.44	$0.87	US	Cheapest latency from US/EU
Fireworks AI	--backend fw	$1.74	$3.48	US	Fastest inference
Anthropic	--backend anthropic	$3.00	$15.00	US	Original Claude Opus

Cost comparison

• Light usage (10 days/month): Anthropic Max $200/mo → DeepSeek ~$20/mo → 90% savings
• Heavy usage (25 days/month): $200/mo → ~$50/mo → 75% savings
• With auto loops: $200/mo → ~$80/mo → 60% savings

DeepSeek's automatic context caching makes agent loops extremely cheap — after the first request, the system prompt and file context are cached at $0.004/M vs $0.44/M uncached.

What works and what doesn't

Works: File reading/writing/editing, bash/PowerShell execution, glob/grep search, multi-step autonomous tool loops, subagent spawning, git operations, project initialization (/init), thinking mode (enabled by default).

Doesn't work or degraded: Image/vision input (DeepSeek's Anthropic endpoint doesn't support images), parallel tool use (disabled — tools execute one at a time), MCP server tools (not supported through compatibility layer). Prompt caching savings are handled by DeepSeek's own system.

Who it's for

Developers running Claude Code heavily on agent loops who want near-identical functionality at a fraction of the cost — especially for iterative coding tasks where DeepSeek V4 Pro's 96.4% LiveCodeBench score is more than sufficient.