Coding agents & IDE setup

MeshTok is OpenAI-compatible and Anthropic-compatible, so every coding agent that supports either format works out of the box. Pick your tool below, set the base URL and your MeshTok key, and you're running on any of 175 models.

Claude Code (Anthropic format)

# Set environment variables, then run claude as normal
export ANTHROPIC_BASE_URL=https://meshtok.com
export ANTHROPIC_API_KEY=sk-your-meshtok-key
export ANTHROPIC_MODEL=anthropic/claude-sonnet-5

claude

Or put them in ~/.claude/.env. Claude Code reads these on every launch. Use any Claude model id from our catalog.

Cursor

Settings → Models → toggle on "OpenAI API Key" → Override base URL:

Base URL:  https://meshtok.com/v1
API Key:   sk-your-meshtok-key

# Then add custom model ids under "Add Model":
#   deepseek/deepseek-v4-pro
#   bigmodel/glm-5.2
#   openai/gpt-5.2
#   anthropic/claude-sonnet-5

Disable the "Cursor" tab/Composer built-in models and use these in Chat and Agent modes. Verify with a "Hello" — it should stream.

Cline (VS Code)

Cline settings → API Provider → "OpenAI Compatible":

Base URL:        https://meshtok.com/v1
API Key:         sk-your-meshtok-key
Model ID:        deepseek/deepseek-v4-pro  (or any of 175)

Cline sends standard OpenAI tool-calling — every MeshTok model tagged Tools works (154 models).

Aider

# OpenAI-compatible models via --openai-api-base
aider --model openai/deepseek/deepseek-v4-pro \
      --openai-api-base https://meshtok.com/v1 \
      --openai-api-key sk-your-meshtok-key

# Or set in environment:
export OPENAI_API_BASE=https://meshtok.com/v1
export OPENAI_API_KEY=sk-your-meshtok-key
aider --model deepseek/deepseek-v4-pro

Codex CLI / OpenAI tools

export OPENAI_API_KEY=sk-your-meshtok-key
export OPENAI_API_BASE=https://meshtok.com/v1
codex --model openai/gpt-5.2

Continue.dev (JetBrains & VS Code)

Edit ~/.continue/config.json:

{
  "models": [{
    "title": "MeshTok DeepSeek",
    "provider": "openai",
    "model": "deepseek/deepseek-v4-pro",
    "apiBase": "https://meshtok.com/v1",
    "apiKey": "sk-your-meshtok-key"
  }]
}

MCP servers & custom agents

Anything that speaks OpenAI's /v1/chat/completions works without modification — set the base URL to https://meshtok.com/v1 and the key as the bearer token. For MCP servers themselves, configure their backing LLM provider the same way.

Recommended models for coding

• Best quality: Claude Sonnet 5, GPT-5.2, DeepSeek V4 Pro
• Best value: DeepSeek V4 Flash, GLM 5.2, Qwen 3 Coder
• Large context: Gemini 2.5 Pro (2M ctx), Kimi K2 (256K ctx)

Browse all coding-capable models on the catalog.