Local-first Memory Wiki for AI agents with hybrid recall, context assembly, and memory management via MCP
MemoryWiki exposes 9 tools with complete input schemas and descriptions. Naming follows verb_noun convention (memorywiki_recall, memorywiki_write_session, memorywiki_forget). Descriptions are present and substantive (80-150 chars typical), explaining WHAT the tool does and WHEN to use it. However, parameter descriptions are generic and lack actionable constraints (e.g., 'Search query for memory recall' does not specify format, length, or example). Output schemas are not documented in the visible code, only input schemas are defined. Error handling is absent from tool definitions; no recovery guidance or categorization. Security is partially addressed via dry_run parameters on write tools, but no explicit permission gates or audit trail documentation visible. Tool composition is sound (each tool has one responsibility), but chaining IDs are not consistently returned (e.g., memorywiki_recall returns results but unclear if they include identifiers needed by memorywiki_read_memory).
Assemble a read-only, metadata-first MemoryWiki context capsule.
Crystallize approved content into semantic/procedural memory.
Dry-run or apply an audited memory deletion.
Check or explicitly rebuild retrieval sidecar indexes.
Dry-run or apply source ingest from the memory root sources/ sandbox.
List MemoryWiki memories by scope, kind, or concept.
Read a specific MemoryWiki memory item or hot file.
Output schemas not documented. Tool definitions show input schemas only; return types and field structures are not visible in the provided code. LLMs cannot plan downstream calls or extract chaining IDs without knowing what fields are returned.
Parameter descriptions lack actionable constraints. E.g., 'Search query for memory recall' does not specify length limits, format, or examples. 'Minimum memory strength threshold' lacks range (0-1? 0-100?). LLMs cannot validate inputs without explicit bounds.
No error handling or recovery guidance in tool definitions. Tools do not document what errors are retryable, user-fixable, or fatal. No guidance on what to do if a memory lookup fails or a write is rejected.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Hybrid recall over local MemoryWiki memory with warnings and provenance.
Save an explicit MemoryWiki session summary. Requires MCP write gates.
Dry-run parameters present but no confirmation/approval workflow documented. Destructive tool (memorywiki_forget) supports dry_run but no explicit confirmation step or approval gate is visible. Agents could accidentally delete memories.
Tool composition risk: memorywiki_recall returns results but unclear if they include identifiers (e.g., memory_id, kind, identifier) needed by memorywiki_read_memory or memorywiki_forget. Broken chaining IDs force extra lookup calls.