Introducing ModelTunnel AI Native Cloud: Better frontier models, faster runs, lower costs →

Claude Code

Claude Code can use ModelTunnel in two ways: as an MCP server that gives Claude access to every ModelTunnel model as a tool, and through ModelTunnel skills.

Add the ModelTunnel MCP server

claude mcp add --transport http deepnexus https://modeltunnel.ai/mcp \
  --header "Authorization: Bearer sk-dn-..."

Inside Claude Code, run /mcp to confirm the server is connected. Claude can now:

  • ask any ModelTunnel model for a second opinion or a cheaper sub-task (chat tool),
  • list models and prices (list_models),
  • check your balance and recent spend (get_balance, get_usage).

For example: "Ask deepseek-v4-pro through deepnexus to review this function for race conditions."

See MCP Server for the full tool reference.

Install ModelTunnel skills

Skills are folders with a SKILL.md that Claude loads when relevant. Install one from the Skills page:

mkdir -p ~/.claude/skills/x402-paywall
curl -fsSL https://modeltunnel.ai/skills/x402-paywall/SKILL.md -o ~/.claude/skills/x402-paywall/SKILL.md

Use .claude/skills/ inside a project instead of ~/.claude/skills/ to share a skill with your team through git.

Use ModelTunnel models with OpenAI-compatible coding agents

Claude Code itself runs on Anthropic models. If you want a terminal coding agent that runs entirely on ModelTunnel models, use ModelTunnel Code or Codex.

Claude Code · ModelTunnel