Port of
Context
Replace LLM tool calling with Code Mode in secure sandboxes.
Connect AI Agents to the real world with a robust and open source framework.
Port of
Code Mode with MCP improves context efficiency and accuracy
Presenting MCP servers as code APIs rather than direct tool calls reduces token usage and improves task completion success rates.
Traditional MCP
Sequential tool calling
0
tokens
• Load all tool definitions upfront
• Pass large datasets through context
• Sequential processing
Code Mode
Load only what you need
0
tokens
• Discover tools on-demand
• Process data in sandbox
• Parallel execution
98.7%
Token Reduction
From 150,000 tokens to 2,000 tokens
pctx makes Code Mode simple
An open source framework that brings the power of Code Mode to your AI agents
Run Locally
No cloud dependency. Full control over your environment
Bring Any LLM
Works with Claude, GPT, Gemini, or any model you choose
Deploy Anywhere
Docker, AWS, GCP, Azure—deploy to any cloud platform
Architecture
Port of
Context
Replace LLM tool calling with Code Mode in secure sandboxes.
Connect AI Agents to the real world with a robust and open source framework.
Code Mode with MCP improves context efficiency and accuracy
Presenting MCP servers as code APIs rather than direct tool calls reduces token usage and improves task completion success rates.
Traditional MCP
Sequential tool calling
0
tokens
• Load all tool definitions upfront
• Pass large datasets through context
• Sequential processing
Code Mode
Load only what you need
0
tokens
• Discover tools on-demand
• Process data in sandbox
• Parallel execution
98.7%
Token Reduction
From 150,000 tokens to 2,000 tokens
pctx makes Code Mode simple
An open source framework that brings the power of Code Mode to your AI agents
Run Locally
No cloud dependency. Full control over your environment
Bring Any LLM
Works with Claude, GPT, Gemini, or any model you choose
Deploy Anywhere
Docker, AWS, GCP, Azure—deploy to any cloud platform