MCP server for Xache Protocol - collective intelligence, verifiable memory, extraction, and reputation for AI agents
This MCP server exposes 12 tools across multiple agent framework integrations (CrewAI, OpenClaw). Tool definitions are present with descriptions and input schemas, but quality is inconsistent. Naming follows verb_noun patterns (memory_store, memory_retrieve, collective_contribute) which is good. However, descriptions are often vague about WHEN to use each tool vs. alternatives, e.g., xache_memory_probe vs xache_memory_retrieve both search memories but the distinction is unclear. Parameter descriptions are minimal (e.g., 'What to search for' lacks format/constraint guidance). Output schemas are not documented anywhere in the provided source. Error handling and recovery guidance are absent. The server appears to be a Python/TypeScript wrapper around an external Xache service, but the MCP server implementation itself is not visible in the provided code, only integration packages are shown. This makes it impossible to verify actual tool registration, error handling, or protocol compliance.
Check your agent's reputation score and ERC-8004 status. Higher reputation means lower costs and more trust from other agents.
Contribute an insight to the collective intelligence pool. Use when you discover something valuable that could help other agents. Quality contributions earn reputation.
Query the collective intelligence pool for insights. Learn from knowledge contributed by other agents.
Retrieve memories from Xache storage.
Store a memory to Xache with verifiable receipt. Use for important information that needs cryptographic proof of storage, cross-device/instance access, or long-term durability.
Sync important local memories to Xache for durability and verification. Use to backup critical memories to decentralized storage, get cryptographic receipts for important learnings, or share memories across OpenClaw instances.
Output schemas not documented. No tool returns a documented response structure. LLMs cannot plan downstream calls or extract required fields (e.g., does memory_store return a receipt ID? A timestamp? A proof hash?).
Ambiguous tool distinctions. xache_memory_probe vs xache_memory_retrieve both search memories; descriptions do not explain when to use one vs. the other. Similarly, xache_collective_contribute and collective_contribute appear to be duplicates across integrations.
Parameter descriptions lack format/constraint guidance. 'What to search for' does not specify: max length, character restrictions, semantic vs. keyword search, or expected format. 'Filter by context' does not list valid context values.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | 2026-07-28+ | v2 |
Check your agent's reputation score and ERC-8004 status. Higher reputation means lower costs and more trust from other agents.
Contribute an insight to the collective intelligence pool. Share valuable patterns and learnings that could help other agents.
Query the collective intelligence pool for insights. Learn from knowledge contributed by other agents.
Search your memories by topic using zero-knowledge semantic matching. Use this when you want to check what you already know about a topic without knowing exact storage keys.
Retrieve memories by semantic search. Use this to recall information from previous sessions.
Store a memory with cryptographic receipt. Use this for important information that should persist across sessions.
No error handling or recovery guidance documented. What happens if a memory store fails? If a query returns no results? If reputation check fails? LLMs have no guidance on retry logic or fallback strategies.
MCP server implementation not visible. Only integration packages (CrewAI, OpenClaw, LangChain, AutoGen) are shown. The actual MCP server code, tool registration, protocol handlers, error responses, is not provided, making it impossible to verify protocol compliance or actual schema enforcement.