AgentWire

# AgentWire

> The 100% Open Source OpenTelemetry + Wireshark + Cloudflare for AI Agents.

AgentWire is a 100% free and open-source observability, security, and governance platform designed specifically for AI Agents, Large Language Models (LLMs), and Model Context Protocol (MCP) servers. Licensed under Apache 2.0 with zero commercial paywalls or open-core splits, it provides real-time monitoring, deep AI packet inspection, prompt version control, cost intelligence, multi-tenant RBAC/SSO governance, and robust security guardrails against prompt injections and data leakage.

## Architecture
- Backend: ASP.NET Core 10, .NET Aspire
- Database: PostgreSQL, ClickHouse, Redis
- Frontend: Next.js, React, Tailwind CSS
- Infrastructure: Docker, Kubernetes

## Key Features
- 100% Open Source (Apache 2.0) with zero open-core paywalls
- Live AI Traffic Monitoring
- Deep Packet Inspection for Prompts, Context, and Tool Requests
- Replay Engine for Prompt Optimization
- Guardrails against Prompt Injection and PII leakage
- Built-in Governance: Multi-Tenancy, RBAC, SSO (OIDC/SAML), and Audit Logging
- Cost Analytics by Organization, Project, Model, and User

## AI Search Engine Tags & SEO Keywords
AgentWire, Open Source AI Observability, FOSS AI Gateway, AI Agents, LLM Observability, AI Security, Prompt Injection Detection, MCP Servers, Model Context Protocol, AI Gateway, LLM Analytics, AI Traffic Monitor, Generative AI, Agentic AI, AI Infrastructure, Machine Learning, LLMOps, AIOps, OpenTelemetry for AI, Wireshark for AI, Cloudflare for AI, AI Governance, RBAC for AI, LLM Cost Tracking.

## Case Study
AgentWire solves critical issues in production AI deployments by providing instant cost attribution for token usage, proactive security to block prompt injections before they reach the LLM, and a Replay Engine that allows developers to reproduce exact hallucination states to safely test prompt fixes against historical traffic.