MCP-compatible context and memory system: paged code navigation for AI assistants, plus a recorded, verified theory of what is known about the code.
Orven provides 11 well-named tools with consistent verb_noun patterns (orven_page_in, orven_list_symbols, orven_search_symbols). All tools have descriptions (avg ~60 chars, within baseline 34-392 range). Input schemas are present and typed for all tools. However, descriptions are terse and lack context about WHEN to use each tool or what downstream operations they enable. Parameter descriptions are minimal (e.g., 'The symbol ID to page in' lacks format/constraint details). Output schemas are not documented, LLMs cannot predict what fields to expect from responses. Error handling guidance is absent. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in metadata.
Get all symbols that call a given symbol
Get the dependency graph for a symbol
Get the L2 skeleton view of the entire page table
Get the current status of the memory system
Get detailed information about a specific symbol
List all symbols in the page table with optional filtering
List all recorded theory statements
Page a symbol body into L1 memory from the L2 skeleton
Output schemas not documented. LLMs cannot predict response structure (fields, types, nested objects). Forces trial-and-error parsing and wastes tokens on clarification.
Descriptions are terse (avg ~60 chars) and lack context. Missing WHEN to use each tool, prerequisites, and how results chain to downstream operations. E.g., 'Get detailed information about a specific symbol' does not explain what 'detailed' means or when to call this vs orven_get_skeleton.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Evict a symbol body from L1 memory back to L2
Record a theory statement about the codebase
Search for symbols by name pattern
No tool annotations despite clear risk metadata. orven_record_theory is WRITE but lacks destructiveHint. All READ_ONLY tools lack readOnlyHint. Agents cannot infer safety/idempotency without annotations.
Parameter descriptions lack format/constraint details. E.g., 'kind' in orven_list_symbols lists valid values in description text ('function, method, class...') instead of as an enum constraint. LLMs cannot parse prose constraints reliably.
No error handling guidance. Tools provide no recovery hints (e.g., 'Symbol not found. Try orven_search_symbols() with a partial name'). Agents cannot self-correct on failures.