Connect any MCP-capable AI to your EREBYX memory substrate. Persistent memory across every AI you use. Native CLI binary; encrypted in transit (TLS 1.3) and at rest with envelope encryption.
erebyx-cli defines 6 tools with partial schema coverage. Tools 1-4 and 6 have explicit input schemas with type definitions and parameter descriptions (good baseline). Tool 5 (update) lacks visible schema entirely. Descriptions are present but brief (10-50 chars), below the 50-200 char LLM-optimized baseline. No output schemas documented. Error handling is absent from tool definitions. Tool naming follows verb_noun pattern (restore_identity, load_context, save, remember, update, wrap_up) but lacks clarity on mutual exclusivity and dependencies (e.g., remember's 'generative' enum with 'auto|true|false|compare' is confusing). Parameter descriptions exist but lack format constraints, ranges, and dependency documentation. No tool annotations (readOnlyHint/destructiveHint) visible despite clear risk classifications in metadata.
Resume from where you left off (WHERE YOU WERE)
Find relevant memories
Load identity at session start (WHO YOU ARE)
Save something that matters
Update an existing memory
Create handoff for session continuity
Tool 'update' has no visible input schema or description in source code. Cannot verify parameter types, constraints, or purpose.
Descriptions are too brief (10-50 chars vs. 50-200 char baseline). 'Load identity at session start (WHO YOU ARE)' lacks WHEN to use, prerequisites, and return value guidance. LLMs cannot reliably select tools with such sparse docstrings.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., what does 'remember' return? Is it a list of memory objects with IDs?). Breaks tool chaining.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Parameter 'generative' in 'remember' uses enum ['auto', 'true', 'false', 'compare']. Mixing boolean strings with semantic values is confusing. Should be ['auto', 'enabled', 'disabled', 'compare'] or clarify in description what 'true'/'false' mean vs. 'auto'.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) in schema despite metadata marking tools as READ_ONLY or WRITE. MCP spec 2026-07-28 expects these hints for agent planning.