Port of Context is the execution layer between AI agents and the tools they call. Tool definitions become typed TypeScript, run in isolated Deno sandboxes. One typed code block replaces dozens of tool calls, with up to 98.7% token reduction.
Port of Context was born over coffee at 7th Avenue Donuts and Diner in Park Slope, Brooklyn. The founders were tired of developer-unfriendly AI agent building infrastructure and set off to build an alternative—for developers, by developers. Today Port of Context powered Goose, Prudential, and others.
Our stack
Built in Rust, with a Python toolkit for converting native agent tools to Code Mode. What the agent sees is TypeScript that we generate from your tool definitions, type-check in under 100ms, and execute inside isolated Deno sandboxes. It speaks MCP, and plugs into LangChain, the Claude Agent SDK, CrewAI, the OpenAI Agents SDK, and Pydantic AI. MIT. Self-hosted. BYO-LLM.
Source: github.com/portofcontext/pctx
Docs: docs.portofcontext.com
0 posts published
1 member
loading...
We're a place where coders share, stay up-to-date and grow their careers.