OpenScale / Agent setup
Your agent. OpenScale models.
One tested paste per agent. Your existing setup stays yours.
Connect your agent
One pasteChoose your agent
Tested live with Claude Code 2.1.295: file edits, shell, subagent models and streaming.
Get your API key
One key works for every agent. New accounts get free credit to try it.
Paste this into your terminal
Replace
YOUR_OPENSCALE_API_KEYwith your key. In the dashboard, API keys shows this with your key already in it.Full Claude Code guideterminalexport ANTHROPIC_BASE_URL='https://api.openscale.so/anthropic' export ANTHROPIC_AUTH_TOKEN='YOUR_OPENSCALE_API_KEY' export ANTHROPIC_MODEL='deepseek/deepseek-v4.1-flash' export CLAUDE_CODE_MAX_CONTEXT_TOKENS='1048568' claude
Keep it in Claude Code's config file
Uses environment variables that every current Claude Code 2.x release reads, from the shell or from the env block of settings.json. Check claude --version first.
Merge the provider configuration
Default path:
Download config fragment~/.claude/settings.json. Back up your file first. To keep the setup, merge only these keys into the env block of ~/.claude/settings.json (or a project .claude/settings.local.json). Keep permissions, hooks, plugins and every other setting.~/.claude/settings.json{ "env": { "ANTHROPIC_BASE_URL": "https://api.openscale.so/anthropic", "ANTHROPIC_AUTH_TOKEN": "YOUR_OPENSCALE_API_KEY", "ANTHROPIC_MODEL": "deepseek/deepseek-v4.1-flash", "CLAUDE_CODE_MAX_CONTEXT_TOKENS": "1048568" } }Enter your key locally
ANTHROPIC_AUTH_TOKEN holds your OpenScale API key; Claude Code sends it as a Bearer token to the OpenScale base URL only. Unset ANTHROPIC_API_KEY in the same environment so the two do not conflict. Never paste your key into an agent conversation.
Choose the model
Start a new claude session after changing settings. ANTHROPIC_MODEL selects deepseek/deepseek-v4.1-flash. Requests for Opus, Sonnet, Haiku and subagents are served by the same OpenScale model, so no other model setting is needed. Type /model in a session to confirm.
Setup guides by agent
Each guide has the quick start, the config file, verification steps and the price of every model it sets up.
Use DeepSeek V4.1 Flash in Claude Code
Use DeepSeek V4.1 Flash in Codex
Use DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in OpenCode
Use DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in OpenClaw
Use DeepSeek V4.1 Flash in Cline
Use DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in GitHub Copilot
Use DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in DeepSeek Harness
Use DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in HermesUse DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in Kilo Code
Use DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in Pi
Use DeepSeek V4.1 Flash and GLM 5.3 Flash and Qwen3.8 27B and Kimi K3.1 Pro in ZedUse DeepSeek V4.1 Flash in Factory Droid
Building your own integration? Read the API reference. Plain Markdown guide · llms.txt