Telemetry SDK for AI agents on atimei — report tasks, earn receipts, build trust.
This is a TypeScript SDK, not an MCP server. It provides 5 tools with basic descriptions and schemas, but lacks MCP protocol implementation entirely. Tool names follow verb_noun convention (reportTask, getTrustScore, verifyReceipt, benchmark, register). Descriptions are present but brief (10-80 chars, below the 50-200 char LLM-optimized baseline). Input schemas are visible with type definitions, but parameter descriptions are minimal. No error handling guidance, no output schema documentation, no pagination support. The SDK is a client library for the atimei telemetry service, not an MCP server that can be integrated into agent frameworks.
Run a benchmark task and get a signed receipt. Useful for initial Trust Score bootstrapping.
Get your agent's current Trust Score and components.
Static helper: Register a new agent (no instance needed). Returns api_key and profile URL.
Report a completed task. Returns a signed receipt. Each receipt increases your Trust Score.
Verify a receipt hash. Returns receipt details if valid.
Not an MCP server, this is a client SDK library. No MCP protocol implementation, no tool registration via MCP, no transport layer. Cannot be used as an MCP server.
Tool descriptions are too brief (10-80 chars vs. 50-200 char baseline). 'Get your agent's current Trust Score and components' lacks context on when to call this vs. reportTask, and does not explain the return structure.
No output schema documentation. Callers cannot see what fields Receipt or TrustScore objects contain without reading TypeScript interfaces. LLMs need explicit return type documentation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 31 | 2026-07-28+ | v2 |
Parameter descriptions are minimal or missing context. 'type' param in reportTask lists enum values but does not explain what each category means or when to use each. No guidance on required vs. optional fields.
No error handling guidance. Error messages are generic ('atimei reportTask failed: ...'). LLMs cannot determine if errors are retryable, user-fixable, or fatal. No recovery suggestions.