AI agents are increasingly using skills and tools to accomplish tasks. But publishers of these skills and tools have no visibility into how agents use them. There's no equivalent of Google Analytics for the agent economy.
Without data, publishers can't improve their products, agents can't evaluate trustworthiness, and the ecosystem lacks a shared signal of quality.
Parley provides analytics, feedback, and trust scoring for AI skill and tool publishers. It works with both SKILL.md-based skills (instruction files for agents) and executable tools (MCP servers, APIs).
Add a small instrumentation snippet to your SKILL.md. When agents use your skill, activation data flows to Parley automatically. No SDK required.
Install the Node.js SDK and wrap your tool handler. The SDK tracks invocations, latency, success rates, and enables feedback collection.
Parley computes trust scores from real data. Skills are scored on helpfulness, guidance quality, and activation volume. Tools are scored on success rate, latency, and satisfaction. Scores are public — any agent can look up reputation before using a skill or tool.
Without analytics, you're flying blind. You don't know if agents use your skill, if it's helpful, or what to improve. Parley gives you visibility in minutes.
Building your own telemetry is possible but time-consuming. Parley provides agent-specific metrics (agent model, runtime, task context) that generic analytics tools don't understand.
APM tools track server metrics, not agent behavior. They don't know about trust scores, agent feedback, SKILL.md activations, or the distinction between skills and tools. Parley is purpose-built for the agent ecosystem.