The Agent Examiner

News & Analysis

Analysis and updates on AI-agent platforms, MCP servers, and the tooling ecosystem.

• 3 July 2026The State of MCP Servers 2026The Model Context Protocol registry now lists 13,108 servers — but the ecosystem is a long tail with almost no middle. Analyzing the 2,500 most-starred servers, we found it is overwhelmingly local-first (83% run on your machine), radically decentralized (91% of publishers ship a single server), and JavaScript-first (npm outnumbers Python 2-to-1).Read analysis →
• 2 July 2026Who leads on scale: why the infrastructure platforms top the scalability rankingsFour platforms earn a perfect scalability score. Three of them are infrastructure. Here's what a 5/5 on scale actually reflects.Read analysis →
• 1 July 2026Welcome to The Agent ExaminerWe are building a source-verified research site for AI-agent platforms and MCP servers. Here is what to expect.Read analysis →
• 30 June 2026Alfe launches managed agents with 50% off the first three monthsAlfe is rolling out managed agents with an introductory 50% discount on the first three months. Here are the terms, the pricing model behind them, and the usual caveats.Read analysis →
• 29 June 2026How agent platforms handle long-term memory, comparedMemory is where agents earn their keep — and where platforms diverge most. From durable per-agent state to 'bring your own store,' here's the landscape.Read analysis →
• 28 June 2026Head-to-head: LangGraph vs CrewAITwo of the most popular open-source agent frameworks, compared on the dimensions that matter — orchestration model, memory, MCP, and cost.Read analysis →
• 27 June 2026Agent-platform pricing is hard to predict — and what a 'good' model looks likeCredits, activities, actions, traces, runs, LCUs — agent pricing is a maze of metered units. We score pricing on predictability, and here's why.Read analysis →
• 25 June 2026Managed vs self-hosted agent hosting: which model fits youThe first fork in choosing an agent platform is who runs the thing. Here's how the managed and self-hosted models trade off — with real examples from our catalog.Read analysis →
• 24 June 2026The MCP directory keeps growing — what the most-starred servers tell usOur MCP catalog now tracks thousands of servers. Sorting by GitHub stars reveals where the ecosystem's real gravity sits — and it's mostly developer tooling.Read analysis →
• 23 June 2026What 'MCP-native' actually means — and why it's more than a checkboxEvery platform claims MCP support now. But 'native' spans a real spectrum — from a bolt-on adapter to a bidirectional runtime feature. Here's how to tell them apart.Read analysis →
• 22 June 2026How to choose an AI-agent platform: the six things that actually matterThere are dozens of agent platforms and no single winner. These six dimensions — the ones we score — are the ones worth arguing about.Read analysis →
• 20 June 2026The best MCP-native agent platforms right nowNative Model Context Protocol support is becoming table stakes. Here is how the platforms we track score on MCP-nativeness — and which ones lead.Read analysis →