AI agent memory & session orchestrator for MCP — persistent KV-Cache, Soul Board, immutable Ledger
n2-soul has 27 tools with moderate schema coverage but significant description and naming gaps. Most tools have basic descriptions (50-150 chars), but many lack actionable guidance on when to use them vs. similar tools. Parameter descriptions are present but often generic. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Schemas are partially visible in KV-Cache tools but absent or inferred for sequence/brain tools. Error handling is minimal, no recovery guidance or actionable error messages visible. Composition is reasonable (single responsibility per tool), but naming could be more verb-forward and consistent.
Read brain/core memory tool registered in registerBrainTools
Write brain/core memory tool registered in registerBrainTools
List all keys for an agent
Load agent core memory
Remove key from core memory
Write key-value to core memory
Batch upsert multiple entities (single disk write for performance)
Get entity by name
Get entities by type
Sequence tools (n2_boot, n2_work_start, n2_work_end, n2_end, brain_read, brain_write) have no visible input schemas. Tool definitions are inferred from registration code, not explicitly shown. Per HARD SCORING RULE, schema score must be 0 for these tools.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications (READ_ONLY, WRITE, DESTRUCTIVE). LLMs cannot infer safety properties from names alone. entity_remove and entity_prune are destructive but unmarked.
Descriptions for sequence tools (n2_boot, n2_work_start, n2_work_end, n2_end) are under 50 chars and lack actionable context. No guidance on when to call them, what they return, or how they compose.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
List all entities
Prune old entities (not mentioned for maxAge days)
Remove entity by name (optionally filter by type)
Search entities by keyword
Add or update an entity (upsert). Merges attributes on match.
Get recent conversation dates for an agent
Read conversation logs for an agent on a specific date
Write a conversation log entry for both caller and target
Backup all project data into a sqlite-store compatible DB
Load the most recent snapshot for a project
Restore from backup
Save a session snapshot with automatic compression to KV-Cache
Search across snapshots by keyword or semantic similarity
Get backup status for a project
Boot sequence tool registered in registerBootSequence
End sequence tool registered in registerEndSequence
Work sequence end tool registered in registerWorkSequence
Work sequence start tool registered in registerWorkSequence
No error handling guidance visible. Tools do not document what errors are retryable, user-fixable, or fatal. No recovery hints (e.g., 'If entity not found, try entity_search() first'). Per pattern:recovery-guide, errors must guide the LLM's next step.
Parameter descriptions are generic and lack format/constraint details. E.g., 'Agent identifier' for 'agent' param does not specify format (alphanumeric, email, UUID?).