The top community-built memory and context management tools for Hermes Agent, ranked by GitHub stars. Memory is what makes Hermes self-improving — these tools extend its built-in recall with long-term semantic memory, graph-based retrieval, and cross-session persistence.
For the architecture behind these tools, read The Hermes Agent Memory Guidebook.
A universal memory layer and official Hermes Agent provider available as a managed cloud or self-hosted OSS. It features multi-level state for users, sessions, and agents using time-aware hybrid retrieval via semantic, keyword, and entity matching.
An official Hermes Agent memory engine featuring a custom vector graph engine with sub-300ms recall and byte-level deduplication. It automates user profile maintenance and integrates with external connectors like Google Drive, Notion, and GitHub.
An AGPL-3.0 context database that uses a hierarchical filesystem paradigm and tiered L0/L1/L2 loading to minimize token costs. It significantly improves task completion rates and provides visualized retrieval trajectories for debugging agent logic.
An opinionated agent brain that synthesizes cited answers across complex ideas and entities. It utilizes a self-wiring knowledge graph for relationship extraction and an automated 'dream cycle' for overnight memory consolidation and gap analysis.
A long-term learning memory system implementing retain, recall, and reflect workflows. It offers a two-line LLM wrapper for easy integration and ensures per-user memory isolation through custom metadata filtering.
A reasoning-first memory library that extracts conclusions from conversations and models users, agents, and projects. It utilizes a hybrid search approach combining BM25 and vector retrieval.
An official Hermes Agent provider that treats memory as a git repository, using a markdown context tree with commit, branch, and merge versioning. It achieves sub-100ms retrieval via a five-tier architecture and includes MCP integration.
A zero-dependency, SQLite-backed memory system delivering sub-millisecond retrieval. It supports the Model Context Protocol for Claude Code and Cursor, featuring temporal knowledge graph triple stores and episodic memory compression.
A recursive self-improving harness that distills successful agent behaviors into local-model training artifacts. It generates persistent playbooks and CLI-first skills specifically formatted for Hermes Agent integration.
A lossless context management plugin that replaces lossy compression with a SQLite-backed DAG engine. It ensures raw message persistence with source-aware lineage and provides tools for searching compacted history.
A self-hosted backend for Honcho that enables private cross-session persistence for Hermes Agent. It supports any OpenAI-compatible LLM or local inference server and includes a Docker-based setup script.
A shared memory layer for multi-agent systems using a readable, open YAML engram format. It provides local-first, data-sovereign storage with zero-cost search and cross-tool compatibility via dedicated SDKs and MCP.
A local-first identity and secrets layer that provides portable state across Hermes Agent, Claude Code, and other MCP clients. It uses SQLite for custody and includes tools to edit, delete, or reclassify context.
An on-device context engine that eliminates cloud dependencies through a local-first architecture. It employs hybrid retrieval combining vector search, BM25, and causal graph traversal for cross-runtime memory sharing.
A context optimization tool that reduces prefill latency and token usage through block reordering and deduplication. It provides native plugin support for the Hermes Agent and OpenClaw ecosystems.
A self-maintaining memory plugin that provides explainable retrieval via 'why_retrieved' tags. It tracks contradictions and maintains hygiene through automated consolidation and a native outcome ledger for agent skills.
A memory kernel stack that consolidates temporal continuity, graph truth, and corpus retrieval. It utilizes a multi-layered storage approach combining SQLite, Kuzu, and Chroma to reduce prompt noise via bounded evidence packing.
An anticipatory memory system that prefetches context by monitoring agent session logs. It uses local-first storage via SQLite-vec and GGUF models, operating as an MCP-native server.
A reflective memory system featuring semantic search, BM25 full-text, and ranked graph traversal. It automates indexing for URLs and codebases and uses edge tags to create bidirectional relationships between notes.
A utility that reduces prompt token overhead by using local BM25 ranking for tool selection. It includes a dashboard to track schema token savings and is compatible with Hermes Agent v0.14.0.
A local-only plugin using a columnar .lbdb embedded graph database with importance-weighted recall. It surfaces high-priority memories first and supports optional GLiNER2 entity extraction.
A pull-model episodic memory plugin that ensures complete record deletion to prevent persistent summaries from haunting future recall. It maintains a trace.jsonl audit log and uses Claude models for on-demand summarization.
A durable cross-session memory provider featuring semantic search and provenance tracking. It automatically mirrors Hermes native memory tools and supports session snapshots with TTL settings.
A plugin that integrates Exabase M-1 for self-organizing long-term memory. It provides tools for manual memory storage and supports configurable query expansion and result reranking.
A comprehensive suite of 17 tools for memory management and knowledge graph traversal. It combines BM25, vector, and graph search methods with automatic context checkpointing during message compression.