Transport-agnostic Model Context Protocol (2026-07-28) server for Fastly Compute with stateless JSON-RPC dispatch, MRTR, and Tasks extension support
Three tools with complete input schemas and descriptions, but significant gaps in parameter documentation and output schema specification. Echo and BookingTool have well-defined inputs; LongJobTool lacks parameter descriptions. No output schemas documented for any tool. Parameter descriptions are minimal or absent (e.g., 'date' in BookingTool has no description). Error handling is implicit rather than explicit. Tool names follow verb_noun convention (echo, book_table, long_job) but descriptions are brief (10-34 chars, below the 50-200 char LLM-optimized baseline). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk profiles (echo=READ_ONLY, book_table=WRITE, long_job=WRITE).
Book a table; asks for a date if missing.
Echo a message back.
Start a long-running job; poll with tasks/get.
No output schemas documented for any tool. LLMs cannot infer what fields to expect or plan downstream tool calls.
Parameter descriptions missing or minimal. 'date' in book_table has no description; LongJobTool input is empty object with no explanation of what it does or what context it needs.
Tool descriptions are too brief (10-34 chars vs. 50-200 char baseline). 'Echo a message back.' and 'Book a table; asks for a date if missing.' lack context for LLM selection and when to use each tool.
No tool annotations despite clear risk profiles. echo is READ_ONLY, book_table and long_job are WRITE operations. Missing readOnlyHint and destructiveHint annotations prevent agents from reasoning about safety and idempotency.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
LongJobTool input schema is empty object with no properties or description. Unclear what context or parameters the tool expects, if any.