EchoMind Memory Engine
jasonatgit/echomind_memory.skill
Local SQLite memory engine for Hermes Agent and other agents, with reflection and forgetting curves
EchoMind Memory Engine is a long-term memory engine for AI agents including Hermes Agent, OpenClaw, OpenCode, DeepSeek Haeness and Claude Code. It stores memories in a local SQLite file and reflects on them to distill rules.
What EchoMind Memory Engine does
EchoMind Memory Engine gives agents long-term memory that remembers preferences, code style, fixes and research methods. It stores seven types of memory, User, Task, Experience, Context, Knowledge, Research and Reflection, in a single SQLite file, with schema migration and transactional writes and no Docker, PostgreSQL or Redis required. A self-evolving engine agent distills semantic knowledge and procedural rules from raw memories and discards low-confidence reflections to limit hallucination pollution.
Stale records are down-ranked or excluded with an Ebbinghaus forgetting curve, and each entry moves through an Active, Stale, Archived and Superseded lifecycle. Retrieval combines keywords, reinforcement-learning weights adjusted by user feedback and LLM semantics. Knowledge entries carry an epistemic mode that separates user-provided facts from inference, and a provenance chain records the origin agent and session. A compact Markdown block can be injected into LLM context, and a root register(ctx) entry point with plugin.yaml kind: exclusive adapts it to Hermes v0.20 and later, compatible with v0.13 to v0.20.
Key features
- Seven memory types stored in one local SQLite file
- Ebbinghaus forgetting curve and an Active, Stale, Archived, Superseded lifecycle
- Self-evolving reflection that discards low-confidence conclusions
- Reinforcement-learning weights adjusted by user feedback
- Epistemic classification that separates user facts from inference
- Markdown memory dossier export and a compact context injection block
When to use it
- Having an agent remember your coding style and past bug fixes across sessions
- Reusing previously fixed bugs and chosen models as recommendations
- Auditing where a stored piece of knowledge came from
Who it is for: Users of Hermes Agent, OpenClaw, OpenCode and Claude Code who want a local long-term memory engine with no external database.
How it fits with Hermes Agent
Lists Hermes Agent first among its supported agents and includes a Hermes v0.20 and later adaptation, but it is a multi-agent memory engine rather than Hermes-only.
FAQ
What is EchoMind Memory Engine?
EchoMind Memory Engine is a long-term memory system for AI coding agents. It extracts knowledge from conversations, stores it in SQLite and uses reflection to turn it into reusable rules.
Does EchoMind Memory Engine work with Hermes Agent?
Yes. The README lists Hermes Agent among its supported agents and describes a Hermes v0.20 and later adaptation with a root register(ctx) entry point. It states compatibility with Hermes v0.13 to v0.20.
Is EchoMind Memory Engine free and open source?
The repository is public, but it has no license file, so default copyright applies. Check with the author before reusing the code.
Similar memory for Hermes Agent
All memoryLocal SQLite memory for coding agents with a Hermes Agent installer, MCP server and code-aware recall
itsXactlY MazemakerLocal semantic memory with a knowledge graph, dream consolidation and MCP access for agents
amanning3390 FlowState-QMDLocal markdown memory server for coding agents that prefetches context before the agent searches
yepyhun BrainstackHermes-native composite memory provider combining profile, session, temporal graph and corpus recall
mlimarenko IronRAGSelf-hosted knowledge graph and RAG system with a 23-tool MCP server for agents, including Hermes
iflytek MemFlywheelFile-native long-term memory that learns after every run, for Pi, Hermes, OpenCode and OpenClaw
Related guides: SOUL.md for Hermes Agent: what it is and how to write one · Run multiple Hermes agents with profiles