271 MCP servers and agent skills in the Knowledge & Memory category, ranked by quality score.
Persistent, self-evolving memory plugin for Claude Code. Background miner extracts durable lessons (decisions, conventions, bug fixes) from completed sessions via Claude Haiku, stores them as human-readable markdown in an Obsidian vault, and assembles query-tailored context briefings at session start. Local-first, no cloud, no API keys. Self-healing install via uv bootstrap shim, `/cortex-doctor`
AI enhancement middleware: precise retrieval replaces full-memory loading (~80%+ token cut), rule-based conflict arbitration with explainable reasons, and one local SQLite shared across Claude Code/Codex/Cursor/ZCode. Local-first, zero cloud, MIT. `pip install memory-arbiter-mcp`
Durable hybrid memory for AI agents. Combines vector search, BM25, temporal retrieval, and optional Memgraph knowledge graph via reciprocal rank fusion. 6 MCP tools: health, search, answer, feedback, graph_query, graph_add_triple. Self-hosted with Qdrant backend. 82.5% strict accuracy on LoCoMo benchmark. `pip install 'mnemostack[mcp]'`
Zero-config MCP server for searchable documentation. Loads markdown into SQLite (default) or PostgreSQL with FTS5/tsvector keyword search and optional pgvector hybrid semantic search.
Persistent memory for AI coding assistants. Local SQLite, no cloud. 16 MCP tools: store, recall, search, relate, consolidate, export, and credential vault (OS keychain). Typed memories with confidence lifecycle (verified/inferred/stale/deprecated), namespaces, knowledge graph, and hybrid BM25 + semantic search via fastembed ONNX. Works with Cursor, Windsurf, Claude, and any MCP client. `pip instal
Persistent memory for AI agents with sectioned entries (project/user/context/etc), semantic search, CLI, and per-project scope. Local JSON, zero config, no server required.
Local-first review-gated sidecar for Markdown workspaces. Indexes notes and incoming artifacts into SQLite, proposes domains, links, memory, and growth cards with evidence, and keeps source files unchanged until explicit review. `pip install linza-mcp`
Dual-memory MCP system (PostgreSQL 17 + pgvector + NVIDIA NIM). 11 tools: search, save, profile, health, auto-repair, stats, tune, dedup, checkpoint, maintenance. Sparse Selective Caching (SSC) with 4-phase progressive retrieval. Bilingual EN/ES. `python scripts/hipocampo_mcp_server.py`
MCP for the [Sovereign AI Blog](https://sovgrid.org), a hands-on engineering log of self-hosted AI on NVIDIA DGX Spark (GB10/SM121A). 44+ articles covering SGLang, Mistral, Voxtral, OpenClaw. Three tools: full-text search, article retrieval, and SGLang config validator for GB10 hardware. Free, no auth, no client-side tracking. `https://mcp.sovgrid.org/self-hosted-ai?ref=awesome-mcp`
Give AI agents read/write access to your Obsidian vault via MCP
Local-first shared memory and coordination layer for AI coding agents: repository evidence, reservations, handoffs, code graph context, and dashboard review backed by PostgreSQL/pgvector.
Let AI tools like Cursor, VS Code, or Claude Desktop answer questions using your product docs. Biel.ai provides the RAG system and MCP server.
MCP-native, local-first memory for coding agents that turns coding sessions into reusable engineering memory: decisions, gotchas, and domain knowledge.
A private, local research "second brain" for Claude: project-aware memory, cited answers from your own library (won't invent what it can't find), linked notes/meetings/ideas via a domain-specific knowledge layer, daily briefs (news + new papers in your field), a live meeting assistant, cross-pollination across your work, and 34 routed agents. A governed layer between you and the AI, with guardrail
Local-first memory engine for AI-agent teams: private/team/project ACL (a hard gate before ranking), associative recall, and federated sync where revocation propagates across nodes. SQLite + FTS5, no LLM required; MCP server + web console included. `pip install 'agent-memory-os[mcp]'`
Self-hosted MCP server for Obsidian with semantic + full-text search over PostgreSQL/pgvector, wikilink graph traversal, atomic note CRUD, OAuth 2.0, and a self-describing vault guide.
MCP server for AI-agent handoffs. Saves client-encrypted WorkBaton checkpoints and WorkStash notes so Codex, Claude Code, Roo Code, and other MCP clients can resume work without passing full chat history.
Compresses AI chat sessions into a typed knowledge graph (~300–800 tokens). Works file-based in Copilot, Claude Code, Cursor, and Gemini CLI without MCP server mode. Supports local/offline extraction via Ollama — no API key required.
Evidence-backed PR contribution advisor. 550+ case studies from 33 repos, 8 tools for merge probability analysis, pattern matching, and contribution strategy. `pip install prgenius-core`
Vault-native, accountable memory for Claude Code and any MCP client. Markdown is the source of truth, no LLM on the Stop path, redaction before every derived store.
Two-tier memory with hot cache (instant injection) and cold semantic search. Auto-promotes frequently-used patterns, extracts knowledge from Claude outputs, and organizes via knowledge graph relationships.
Self-hosted mem0 MCP server for Claude Code with Qdrant vector search, Neo4j knowledge graph, and Ollama embeddings. Zero-config OAT auth, split-model graph routing, session hooks for automatic cross-session memory, and 11 tools. Supports both Anthropic and fully local Ollama setups.
Git-aware, evidence-based memory store for AI coding agents. Memories anchored to git blob hashes are automatically flagged for review when linked code changes. `pip install robo-cortex[mcp]` or on registry.modelcontextprotocol.io
AI-native structured memory for coding agents — typed lore/doc entries with status and links, read and written by agents via MCP.
Hosted agent memory that keeps an audit trail. Three tools — `commit_mem`, `recall`, `status` — where each write carries provenance (which run, tool, or operator produced it), contradictions with prior state surface as inspectable changes rather than silent overwrites, and a bad run is recovered from a snapshot instead of repaired by hand. `pip install -U bilinc`
Read-only doctrine access for the Agentic AI Blueprint — the industry standard reference for safe, observable, and steerable AI agent UX. Browse and search 10 principles, clusters, curated implementation examples, and application guides for production agentic AI systems. 13 public tools require no credentials. `https://aidesignblueprint.com/mcp`
Persistent knowledge layer for AI agents. KB, investigation threads, journal with multi-agent attribution. SQLite/PostgreSQL/Supabase.
Inspectable memory substrate for Claude Code. Three read-only MCP tools (search_pairs, compact_session, substrate_info) over a local-first, signal-scored parse of `~/.claude/projects/`. Per-pair scores are numpy columns on disk, not opaque vectors in a service. Zero outbound calls (CI-enforced).
Read-only access to the AI Loop Library catalog: 63+ bounded, verifiable agent work loops. `pick_loop_for_goal` recommends a loop for a stated goal and `render_run_protocol` emits an executable markdown protocol with verification, stop conditions, budgets, and approval gates. Single-file, stdlib-only stdio server.
MCP server for OWL/TTL/RDF ontologies. Supports SPARQL queries, HermiT/Pellet reasoning, and CRUD operations on knowledge graphs. Works with Claude Desktop, Cursor, and Windsurf.
Local FAISS vector database for RAG with document ingestion (PDF/TXT/MD/DOCX), semantic search, re-ranking, and CLI tools for indexing and querying
Shared versioned state store with optimistic concurrency control for coordinating concurrent AI agents. SQLite-backed claim/release locks and append-only audit log.
A still point for AI agents in fog. When an agent loses its thread, Oasis distills its original purpose and returns clarity. Pay-per-use with Lightning or Arbitrum — karma discounts via ARGENTUM reputation. Self-hosted with phoenixd.
A Model Context Protocol (MCP) server that implements the Zettelkasten knowledge management methodology, allowing you to create, link, and search atomic notes through Claude and other MCP-compatible clients.
AI-powered paper annotation MCP server. Reads papers, highlights key findings with semantic color coding, explains formulas, and writes structured reading notes — all saved back to Zotero. Features two-phase workflow for large PDFs (63–80% context savings) and batch annotations.
Persistent 4-tier AI memory (episodic, semantic, project, procedural) with temporal scoring, contradiction detection, entity tracking, and real-time desktop visualization orb.
Production-grade persistent memory service for AI agents. Semantic search with recency decay (configurable α blend), local ONNX embeddings (zero API cost), background memory summarisation, and Prometheus/Grafana observability. `docker compose up` — no cloud dependencies.
Read Markdown with inlined images and heading indexes for agentic RAG over local docs or allowed URLs.
Persistent fractal graph memory for AI coding agents via MCP. Semantic vector search, hot/cold/archive context tiers, decision tracking, and Docker-based deployment with Qdrant.
Self-hosted memory and governance layer for AI coding agents. 28 MCP tools with structured knowledge capture, hybrid search (semantic + BM25 + cross-encoder reranking), behavioral documentation nudges, cold-start codebase analyzer, and git-native storage. Single Docker container, zero cloud dependencies.
Agent-first cross-platform bookmark & knowledge manager. AI agents collect web pages from any source (pairs naturally with Fetch, Firecrawl, Brave Search) and auto-categorize into organized collections with search, tags, and shared folders. 16 tools including `agent_register` (zero-config onboarding), `add_footprint`, `search_footprints`, `create_shared_category`. `pip install -r requirements.txt
A Model Context Protocol server for Mem0 that helps manage coding preferences and patterns, providing tools for storing, retrieving and semantically handling code implementations, best practices and technical documentation in IDEs like Cursor and Windsurf
Persistent identity architecture for AI agents. 16 MCP servers covering drives, emotional relationships, semantic memory with decay, working threads, learned patterns, journal, genesis (identity discovery), creative collision engine, forecasting, and voice. Zero dependencies beyond Python 3.8. Built across 938 conversations.
Persistent, searchable context storage across Claude Code sessions using SQLite FTS5. Save sessions with AI-generated summaries, two-tier full-text search, checkpoint recovery, and a web dashboard.
Local workspace memory for Claude Desktop. Indexes your documents (Markdown, CSV, session logs) into a vector store with hybrid search, cross-session memory, auto-learn, and knowledge graph visualization. Zero external dependencies — fastembed + LanceDB, no Ollama or Docker required. 15 MCP tools.
Persistent memory MCP server with OS-style demand paging, kswapd-style eviction, and mlock pinning. Multi-agent shared SQLite store with BM25 + FTS5 retrieval, importance scoring, and capacity-aware reclaim. 5 tools: `memory_lookup`, `pin_memory`, `unpin_memory`, `memory_stats`, `list_pinned`. `uvx --from git+https://github.com/soolaugust/0CompactMem memory-os`
Enable AI assistants to conduct structured, persona-driven sessions including interview preparation, personal reflection, and coaching conversations. Built-in timer management and performance evaluation tools.
Governed memory for coding agents with trust lifecycle (hypothesis → active → validated → deprecated), conflict detection, staleness tracking, and health scoring. SQLite + FTS5, zero infrastructure. `pip install quilmem[mcp]`