Self-hostable Remote MCP server for shared AI memory and artifact recall backed by user-owned storage.
Server has 14 tools with explicit Zod schemas and descriptions. Naming is verb-first and clear (add_memory, search_memory, delete_memory, etc.). However, parameter descriptions are sparse, many lack detail on constraints, formats, or when to use them. Output schemas are not documented; responses are JSON-stringified without field-level guidance. Error handling is minimal; no recovery guidance or actionable error messages visible. Security: no evidence of permission gates on destructive tools (delete_memory, delete_artifact, hard_delete). Artifact tools have richer schemas (enums for artifact_type, provider) but still lack per-parameter descriptions in several cases.
Register a large externally stored log and keep only summary/selected chunks in the database.
Soft-delete artifact metadata by default; optionally hard-delete and remove storage object.
Extract useful text chunks and summary from an artifact.
List registered artifacts by project, type, provider, or tag.
Memory tools (add_memory, search_memory, update_memory, delete_memory, export_memories) lack visible input schemas in source. Only artifact tools show explicit Zod schema definitions. Cannot verify parameter types, constraints, or descriptions for 6 of 14 tools.
Destructive tools (delete_memory, delete_artifact with hard_delete=true) have no permission gates, confirmation steps, or dry-run support. An agent can irreversibly delete data without authorization checks or user confirmation.
Output schemas are not documented. Tools return JSON-stringified responses with a generic 'ok' + 'data' wrapper, but field-level structure is not specified. LLMs cannot plan downstream calls or extract specific fields reliably.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2025-06-18+ | v2 |
Recall only relevant artifact chunks and optional summary.
Register metadata for a large file/log already stored outside the memory database.
Search artifact metadata, summaries, and extracted chunks.
Report memory database and artifact storage provider status without exposing secrets.
Error handling is absent. No visible error recovery guidance, categorization (retryable vs fatal), or actionable error messages. Stack traces or generic 500 errors will not help agents self-correct.
Artifact tools have enums (artifact_type, provider, indexing_mode) but lack descriptions for enum values. LLMs cannot infer when to use 'summary_only' vs 'selected_chunks' vs 'full_text_if_small' without guidance.