Guide

Use DeepSeek with OpenClaw

DeepSeek offers the cheapest reasoning-capable LLM. This guide shows you how to get an API key, configure DeepSeek in OpenClaw, and pick the right model for your use case.

Quick Answer

To use DeepSeek with OpenClaw: 1) Get an API key from platform.deepseek.com (add $5 credit). 2) Open ClawSquire → Configuration → expand "models" → set provider to DeepSeek and paste your API key. 3) Choose deepseek-chat for best cost/performance. 4) Test via openclaw chat or your Telegram bot.

Why DeepSeek?

💰

Cheapest Reasoning

~$0.14/M input tokens for deepseek-chat

🧠

Strong Performance

Competitive with GPT-4 on reasoning benchmarks

🌐

OpenAI-Compatible API

Works with any OpenAI-compatible client

1

Get a DeepSeek API Key

Go to platform.deepseek.com and create an account. Navigate to API Keys and click Create new secret key.

Copy the key (starts with sk-). You'll need to add credit — $5 is enough for extensive testing.

💡 Pricing tip

deepseek-chat costs ~$0.14/M input tokens and ~$0.28/M output tokens. A typical conversation costs less than $0.01. $5 can last weeks of regular use.

2

Configure DeepSeek in OpenClaw

★ Recommended: Use ClawSquire

ClawSquire makes provider setup visual — no terminal or JSON editing needed.

  1. Open ClawSquire → Configuration page
  2. Expand the "models" section → set provider to DeepSeek and paste your API key visually
  3. Alternatively: the Channels page shows all configured providers at a glance
ClawSquire Configuration page with collapsible tree showing meta (2), models (1), commands (4), gateway (2) — note that secrets (API keys, tokens) are partially masked for security, with JSON View toggle button visibleClawSquire Channels page showing deepseek:manual (api_key) as Configured — all LLM providers and their status at a glance

Manual Alternative

If you prefer the command line, use these CLI commands:

Create an environment file for your API key:

echo "DEEPSEEK_API_KEY=sk-your-key-here" >> ~/.openclaw/.env

Then set the default model:

openclaw config set agents.defaults.model deepseek/deepseek-chat
3

Choose the Right Model

ModelBest ForCost
deepseek-chatGeneral conversation, reasoning$0.14/M in
deepseek-reasonerComplex reasoning, math, code$0.55/M in

For most users, deepseek-chat provides the best balance of capability and cost.

4

Test the Connection

After configuration, verify everything works:

# Via CLI
openclaw chat "Hello, what model are you?"

# Or simply message your Telegram bot

The model should respond and identify itself. If you see errors, check the troubleshooting section below.

⟩Troubleshooting

"Billing error" or "insufficient balance"+

Your DeepSeek account needs credit. Go to platform.deepseek.com → Billing → Add funds. $5 is enough for extensive testing.

"No API key found for provider deepseek"+

The key needs to be in your auth store. Check that ~/.openclaw/.env contains DEEPSEEK_API_KEY and restart the gateway.

Model responds but quality is poor+

Make sure you're using deepseek-chat (not an older model). Check your system prompt isn't too restrictive.

Other LLM Providers

OpenClaw supports multiple providers. Here's a quick comparison:

ProviderBest ForFree Tier?
DeepSeekCheapest reasoningNo (but very cheap)
GeminiFree tier availableYes (15 RPM)
OpenAIImage understandingNo
AnthropicLong context, codingNo

Next Steps

Download ClawSquire — FreeGet DeepSeek Connected — from $199

Guided setup, health diagnostics, 7 languages.