// the blog
Field notes on agent readiness
How AI agents choose and use tools — measured against real models, servers, and verticals, then written down.
-
How LLMs Choose Which Tool to Call (And Why Yours Gets Ignored)
When an agent can reach many tools, what makes it pick yours — and what makes it ignore you? 4,914 real tests across 5 models and 5 verticals.
-
MCP Resources vs. Tools vs. Prompts: When to Use Each
MCP has three primitives — tools, resources, and prompts. What each one does, when agents actually use them, and why the distinction matters.
-
OpenAPI to MCP: How to Make Your API Agent-Ready
Converting an OpenAPI spec to an MCP server is the easy part; making it work well for AI agents is where most conversions quietly fail.
-
MCP vs. A2A vs. Tool Calling: Which Protocol Should You Use?
MCP, A2A, and native function calling solve different problems at different layers — with data on what actually happens at the tool-calling layer.
-
Datadog MCP Server: Lessons Validated Across 17 Others
Datadog's four MCP interventions, measured against the same patterns across 17 servers and 6,265 agent-tool interactions — where they generalize, and where they don't.
-
Agent Readiness Isn't One Thing: Codebase vs. Tool Surface
Agent readiness is two completely different things — codebase vs. tool surface — and confusing them optimizes the wrong half of your stack.