AI Agent platform for incident triage and project Q&A on OpenHands — Git sync, logs, config center, Ingress/Webhook
DeepTicket's demo MCP server has 6 tools with basic schemas and descriptions, but significant quality gaps. All tools have descriptions (10-100 chars, mostly adequate) and input schemas with type definitions. However, naming lacks action verbs (demo_lookup, demo_read_config, demo_echo are vague about intent), parameter descriptions are minimal (1-2 sentences), and output schemas are completely undocumented. Error handling is absent, no guidance on retryability, user-fixable vs fatal errors, or recovery paths. The tools appear designed for testing/demo purposes rather than production use, which explains the generic naming and sparse documentation. No tool annotations (readOnlyHint, destructiveHint) despite clear risk classifications in metadata.
Simulate a destructive action; never deletes anything (dry_run only).
Echo inputs back as JSON — useful for distinguishing multiple tool calls in a timeline.
Simulate shell execution; does not run the command.
Look up a fixed demo record by key (e.g. order-123). Deterministic per key.
Return a fixed in-memory config snapshot; section filters top-level keys when not 'all'.
Optional tool for unknown_tool_mode tests (enable via DEMO_EXTRA_TOOL=1).
No output schemas documented. LLMs cannot predict return structure, forcing them to guess at response fields and plan downstream calls blindly.
Tool names lack action verbs. 'demo_lookup' does not clearly indicate GET vs SEARCH; 'demo_read_config' vs 'demo_get_config'; 'demo_unknown_action' is opaque. LLMs cannot infer intent from names alone.
No error handling guidance. Tools like demo_delete_resource and demo_exec_command have no documented error cases, retryability, or recovery paths. LLMs cannot distinguish transient from fatal failures.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | 2026-07-28+ | v2 |
No tool annotations despite clear risk metadata. demo_delete_resource and demo_exec_command are marked DESTRUCTIVE/IRREVERSIBLE but lack destructiveHint or idempotentHint annotations in schema.
Parameter descriptions are minimal (1-2 sentences). 'The key to look up in the demo fixture' lacks format, range, or example constraints. LLMs cannot validate inputs before calling.