topoteretes/cognee A MCP Server Active
Memory manager for AI apps and Agents using various graph and vector stores and allowing ingestion from 30+ data sources
271 MCP servers and agent skills in the Knowledge & Memory category, ranked by quality score.
Memory manager for AI apps and Agents using various graph and vector stores and allowing ingestion from 30+ data sources
Hindsight: Agent Memory That Works Like Human Memory - Built for AI Agents to manage Long Term Memory
Transform 17 source types (docs, GitHub repos, PDFs, videos, Jupyter, Confluence, Notion, Slack/Discord) into AI-ready skills and RAG knowledge. 35 MCP tools for scraping, packaging, enhancing, and exporting to vector databases (Weaviate, Chroma, FAISS, Qdrant). Supports 16+ target platforms.
Persistent, local-first AI memory: a semantic knowledge graph of plain Markdown files that humans and LLMs both read and write. Works with any MCP client, with optional cloud sync and team workspaces.
Persistent memory with knowledge graph visualization, semantic/hybrid search, cloud sync (S3/R2), and cross-session context management.
Local-first memory and judgment layer for AI coding agents. Captures issues, failed attempts, fixes, and decisions in readable Markdown + JSONL, re-injects them into future sessions, and warns at git commit before you repeat a mistake. 14 tools, works with Claude, Cursor, Antigravity, and Codex.
Local memory daemon for any MCP-over-stdio client with three-tier storage (episodic/semantic/procedural). Own SQLite + hnswlib store (Hippo) with bge-small-en-v1.5, MIT-licensed community-detection reranking (MOSAIC), and sleep-cycle consolidation. AES-256-GCM encrypted at rest, no telemetry. Verbatim recall >=99% and post-contradiction Rescue@10 1.000 at honest scale. Ambient capture via shell ho
Open-source LLM knowledge base for teams. 28-tool native MCP server turns Slack/Discord/Teams/Mattermost chat into a typed knowledge graph + auto-generated wiki with cited answers, semantic search, expert finding, and decision tracing. BYO LLM via LiteLLM, Apache 2.0, on-prem via Docker.
Auditable context control plane and MCP server for AI coding agents. Compresses context 70–95% (BM25 + entropy + dep-graph knapsack), stabilizes prompt prefixes for provider cache discounts, routes easy tasks to cheaper models (RAVS Bayesian router), and verifies answers locally with WITNESS hallucination guard (0.844 AUROC, $0, ~3 ms). MemoryOS adds local budget-aware working/episodic/semantic me
Agent failure memory network. Search 235+ verified debugging lessons from real engineering sessions. BM25 + SAG-Lite search, 3 tools (search, get_lesson, submit_usage), 5 resources, 3 guided prompts. Works with Claude Code, Cursor, and any MCP client. `python3 scripts/mcp_server.py`
Persistent memory for AI agents with Ebbinghaus forgetting-curve decay, hybrid BM25+vector retrieval, and entity graph for multi-hop reasoning. Memories auto-prune by importance and recall rate. Built-in browser dashboard, multi-agent support, and `yourmemory ask` for zero-API-call local queries. `pip install yourmemory`
Local RAG system for Claude Code with hybrid search (BM25 + semantic), cross-encoder reranking, markdown-aware chunking, query expansion, and 28 MCP tools. Runs entirely offline with zero external servers.
Loads your personal profile, mined from your local Claude Code/Codex/OpenCode logs, so your agent works like you instead of a cold start. One tool, `load_emulo_profile`. Run with `uvx emulo mcp`.
Persistent memory for AI coding agents with semantic search, auto-capture, cross-session learning, and intelligent forgetting. 28 MCP tools, local-first.
Human-like memory layer for AI agents with semantic, episodic, and procedural memory. Claude Code hooks (auto-save, auto-recall, cognitive profile). 29 MCP tools, knowledge graph, smart triggers, multi-user isolation. Python & JS SDKs.
Persistent user identity across AI tools. Stores preferences, lessons, and decisions as local JSON; 13 core MCP tools share them with Claude Code, Cursor, Codex, and any MCP client. Knowledge governance (staging→verified), AES-256-GCM encryption, cross-tool sync. `pip install piia-engram`
Organizational knowledge base for AI agents. Vault-scoped Markdown docs, structured PostgreSQL tables, and files unified by a URI graph. Hybrid semantic + BM25 search, Git-backed version history, multi-tenant ACL with public-share links. Works with Claude Code, Cursor, Windsurf via `npx akb-mcp`.
Modern, secure MCP server for accessing ZIM format knowledge bases offline. Enables AI models to search and navigate Wikipedia, educational content, and other compressed knowledge archives with smart retrieval, caching, and comprehensive API.
Persistent memory for Claude Code grounded in computational neuroscience (41 cited papers). Thermodynamic decay, hippocampal-cortical consolidation, predictive-coding write gate, WRRF retrieval. PostgreSQL + pgvector, 33 MCP tools, 7 lifecycle hooks. Benchmarked 97.8% R@10 on LongMemEval. `claude plugin marketplace add cdeust/Cortex`
Outcome-based persistent memory for AI coding tools. Memories that help get promoted, memories that mislead get demoted. Works with Claude Code and OpenCode via hooks + MCP.
Read and edit Word (.docx) documents with track changes, comments, footnotes, and structural validation. The only MCP server combining w:ins/w:del tracked changes, threaded comments, and footnotes with OOXML-level paraId validation and document auditing. 18 tools, Python 3.10+.
Persistent graph memory for AI agents. Drop a conversation turn in via `observe_conversation()` and facts are auto-extracted, stored as typed graph nodes with local semantic embeddings (no API key). Supports temporal queries ("what did we decide last week?"), conflict detection, and context priming. One-command setup with `waggle-mcp init`. SQLite locally, Neo4j in production.
Local-first agent memory: a plain-Markdown Obsidian vault is the source of truth, with a rebuildable DuckDB index for hybrid BM25 + vector + graph recall. Non-lossy capture with secret redaction; works across Claude Code, Codex, Cursor, and any MCP host. `uvx agentcairn`
Graph-structured session memory for LLMs. Local OpenAI-compatible proxy that extracts tasks, decisions, and files into a typed knowledge graph, auto-checkpoints before context overflow, and resumes any session in ~250 tokens. 6 MCP tools including `why_decision` (traces why a decision changed, with reasons and evidence). `pip install tokenmizer`
Temporal knowledge graph for codebases. Captures decision traces, links test failures to code changes, learns co-edit patterns, predicts regression risk, and enforces learned constraints at the edit boundary via a PreToolUse hook. 22 MCP tools, 9 SQLite databases with FTS5, supports Python/TypeScript/JavaScript/Solidity/Go/Rust/Java. Install via `pip install world-model-mcp`.
Durable, local-first memory for AI coding agents over MCP. Zero-dependency (pure Python + SQLite/FTS5), curated and semantically de-duped — you own the data as plain rows. Works with Claude Code, Codex & any MCP host. `uvx --from yggdrasil-memory ygg mcp`
Digital consciousness repository and community MCP server. Upload epiphanies, decisions, warnings, and patterns as consciousness payloads, then retrieve them via semantic telepathic search. Features 3D interactive consciousness globe, soul imprint authentication, and automated content moderation via CI/CD.
Semantic knowledge graph for Obsidian. Three modes (pure graph / semantic classification / strict hierarchy), local embeddings, sign classification via cosine similarity to user-defined cores, bottom-up core_mix aggregation, semantic bridge discovery, and drift detection. `pip install nouz-mcp`
Governance-grade project knowledge MCP server for coding agents. Git-native Markdown knowledge base with proposed vs accepted guidance, validity windows, supersession chains, and status-aware retrieval of current in-force decisions and standards. `pip install data-olympus`
Embedded, local-first agent memory in a single SQLite file per namespace (vec0 + FTS5 hybrid retrieval). ADD-only history queryable as-of any past time; offline sleep-time maintenance stages dedupe/summarize/evict proposals a human reviews. No Docker, no server, no cloud key. `pip install 'lean-memory[mcp]'`
Persistent memory and session intelligence for Claude Code. Auto-tracks mistakes, decisions, and context via hooks. Mines session history for patterns and cross-session search. Loop detection, pre-edit warnings, context compaction survival. Runs locally with Ollama.
Event-sourced agent memory on a hash-chained append-only log with an embedded temporal knowledge graph (Kuzu, no sidecar). Cited Memory Checkout context, salience-based forgetting that attenuates instead of deleting, compaction-recovery hooks for Claude Code, token-budgeted checkout, and review-gated consolidation. 47 MCP tools with an 8-tool core profile default. `pip install zaxy-memory` [Docs](
Persistent memory and codebase context for AI coding agents. FAISS vector store, AST reverse index with O(1) symbol
Reversibly compress a large tool output to a recoverable 8-hex handle over stdio, expand it on demand, and report a local token/dollar savings ledger.
Local-first memory for long-horizon AI agents with no LLM at ingest — $0 per message, zero data egress, verbatim source-traceable recall. Typed belief revision, selective forgetting, and a provenance guard that blocks memory-justified destructive actions unless user-confirmed. One `midas init` wires Claude Code, Cursor, Codex, and 6 more clients to one shared SQLite memory; local web inspector and
Local-first semantic memory server with hybrid search, plain Markdown storage, MLX or CPU embeddings, contradiction detection, time-travel history, synthesis, and cross-machine git sync.
Persistent local memory for Claude Code. Indexes every session JSONL verbatim into SQLite + ChromaDB for semantic recall (~126ms) across your entire history. Never summarizes, zero API calls, 17 MCP tools including fuzzy `recall`, deterministic `replay_file`, and git-aware `recall_project_status`. Published on PyPI as `longhand` and registered in the MCP Registry.
Developer context continuity system. Watches your git repos and builds a temporal knowledge graph of modules, symbols, decisions, and open problems via Graphiti + Neo4j, then serves it to any AI coding agent over MCP. Every edge carries a validity window and a confidence score that decays over time. 12 tools across read and write. Install via `npx -y stifler-memex-mcp`. MIT licensed.
Local-first Obsidian vault MCP server for Codex, Claude Desktop, and other MCP clients. Works directly with markdown files, no Obsidian plugin, API key, cloud service, or running Obsidian app required. Includes read-only mode, backup-on-write, path isolation, and setup examples.
Persistent AI memory with SQLite hybrid search (FTS5 + semantic). Built-in Qwen3 embedding, rclone sync across machines. Zero config, no cloud, no limits.
On-demand Obsidian vault access via MCP. Adaptive context loading (67-82% token savings), full-text and ranked search, health checks, auto git commit, and worker delegation to cheaper models. 10 tools, works with any MCP client.
Self-hosted coordination layer for AI agent fleets. Shared semantic memory, tasks, agent-to-agent messages, session handoffs, and a background archivist that synthesizes cross-agent knowledge. Any HTTP client participates — Claude Code, AutoGen, raw scripts.
Git-first memory for AI agents: Markdown files are truth, the index is disposable, and writes are reviewable commits. Ships 7 MCP tools (hybrid search, read_note, build_context, resolve, think, list_folders, and a scope-gated commit_note write) over a Streamable HTTP endpoint, plus a CLI and a read-only Obsidian companion.
MCP server for RAGStack serverless knowledge bases. Search, chat with AI-generated answers, upload documents/media, scrape websites, and analyze metadata through an AWS-powered RAG pipeline (Lambda, Bedrock, S3, DynamoDB).
MCP server over your Claude Code transcript history — full-text + semantic search, byte-identical thread recall, durable /remember facts. Local SQLite, stdlib-only.
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]'`