pctx
Verified Safeby portofcontext
Overview
Aggregates multiple Model Context Protocol (MCP) servers into a single endpoint, providing a sandboxed TypeScript "Code Mode" interface for AI agents to interact with tools efficiently, reducing LLM token usage.
Installation
pctx devEnvironment Variables
- STRIPE_MCP_KEY
- GRAFANA_CLOUD_TOKEN
- OTEL_USERNAME
- OTEL_PASSWORD
- OTEL_API_KEY
- NASA_API_KEY
- NASA_MCP_PORT
- PCTX_PORT
Security Notes
The system explicitly leverages Deno's secure runtime for executing LLM-generated TypeScript code. It enforces strict sandboxing, limiting network access to only configured upstream MCP hosts and completely disallowing filesystem, environment, or system access. Authentication credentials for upstream MCPs are managed securely by pctx using environment variables, system keychain, or external commands, preventing LLMs from direct access. The architecture demonstrates a strong focus on minimizing the attack surface and securing AI agent interactions.
Similar Servers
mcp-use
A full-stack framework for building Model Context Protocol (MCP) servers, MCP clients, and AI agents in both Python and TypeScript, supporting interactive UI widgets and robust debugging.
ncp
A unified Model Context Protocol (MCP) orchestrator that enables semantic tool discovery, management, scheduling, and execution across a diverse ecosystem of connected AI tools and services.
mcpc
A framework for building agentic Model Context Protocol (MCP) servers by composing existing MCP tools. It enables the creation of portable, interoperable AI agents with flexible execution modes and robust logging/tracing capabilities.
MCP-Agent
An AI agent for discovering, connecting to, and interacting with Model Context Protocol (MCP) servers and their provided tools, resources, and prompts.