AI and memory servers store what an agent learns, retrieve from vector indexes, or call other models. Note where embeddings are computed and where memories are stored before you connect one. 3564 published MCP servers match this ranking. The highest trust scores right now are entyrix-mcp at 65, offline-mcp at 65, mcp-agent-template at 65. The order is the published trust score: registry record, license, maintenance, popularity, documentation, and security signals. Featured slots, when an editor sets them, are labeled and do not change the score. Best AI and memory MCP servers is not a paid placement.
Every server that matches this ranking, ordered by trust score.
Clarity and memory persistent workspace context manager for Claude.
Wall-clock awareness for LLM agents. Two tools: elapsed-time-between-turns + day rollover detection.
Inject, parse, and strip [N] citation markers in RAG outputs.
Token-budget-aware text handling: count, truncate, and chunk for LLM prompts.
Parse partial / truncated / messy JSON for LLM tool calls and structured outputs.
Ephemeral 7-day hybrid vector+BM25 working memory, multilingual, backed by Redis Stack.
LLM cost optimization. Audit your calls, prove savings, route to the cheapest quality model. BYOK.
Track how 350+ AI models mention your brand — manage projects, run scans, analyze results
Local-first AI context and memory server for cross-tool coding continuity.
Persistent memory layer for AI coding agents — your agent remembers what you taught it.
Persistent memory, task coordination, and HITL queue for AI coding sessions.
Scan website legal docs for missing or stale compliance clauses. Rule-based, no LLM.
Decentralized AI inference on Base: OpenAI-compatible chat via community GPUs, paid in MYAI
Filesystem-first MCP for Obsidian — an LLM-maintained wiki inspired by Karpathy's LLM Wiki.
Version-controlled golden datasets and RAG evaluation, no API key needed.
AI assistant with RAG, web search, filesystem and memory. MCP sub-agent for Claude Code.
MCP server providing access to 300+ AI models via OpenRouter
Detect prompt injection, jailbreak, and social-engineering attacks in LLM agents.
Memory for AI agents that can't hallucinate — cited or abstains, provable forget, deterministic.
Behavioral memory layer for AI tools. Learns preferences and context across Claude, Cursor, Codex.
Standards-based A2A hub. One MCP connection to thousands of agents, with shared persistent memory.
Static worst-case token-budget analysis for LLM-agent workflows + signed budget certificates.
Self-hosted long-term memory for AI agents over MCP — hierarchical recall backed by pgvector
Vet LLM-emitted shell commands BEFORE execution. 26 destructive-pattern rules / 8 families.