Mnemosyne Hermes Memory Providers - Setup Guide¶
Source: mnemosyne-oss/mnemosyne GitHub: github.com/mnemosyne-oss/mnemosyne (2.3K⭐) Category: Memory / Knowledge Infrastructure First Seen: July 14, 2026 Security: Gen Agent Trust Hub Pass · Socket Pass · Snyk Pass
Mnemosyne is a local-first memory layer for AI agents. Deployed as a Hermes memory provider, it replaces the built-in MEMORY.md/USER.md flat-file system with SQLite-backed vector + FTS5 hybrid search, episodic consolidation, temporal knowledge graphs, and optional bidirectional sync.
100% local. Zero cloud. Sub-millisecond recall.
Installation¶
npx skills add https://github.com/mnemosyne-oss/mnemosyne --skill hermes-memory-providers
The provider ships with its own SQLite storage engine; no external vector database or cloud account is required.
What It Gives You¶
| Capability | Detail |
|---|---|
| System prompt injection | # Mnemosyne Memory context block in every prompt |
| Pre-turn prefetch | Relevant memories injected before each LLM call |
| Post-turn sync | Conversation turns auto-stored to episodic memory |
| 20 injected tools | remember, recall, sleep, triples, scratchpad, graph, sync, diagnostics, and more |
| 3 lifecycle hooks | pre_llm_call, on_session_start, post_tool_call |
| CLI | hermes mnemosyne {stats|sleep|inspect|export|import|clear|version} |
CLI Reference¶
hermes mnemosyne stats # memory counts, storage size, health
hermes mnemosyne inspect # browse stored memories
hermes mnemosyne export # portable JSON export
hermes mnemosyne import # restore from export
hermes mnemosyne sleep # consolidation cycle (vector re-index)
hermes mnemosyne clear # wipe memory store
hermes mnemosyne version # provider version
Prerequisites¶
| Requirement | Details |
|---|---|
| Hermes Agent | Current version with memory-provider support |
| Python 3.9+ | For the provider runtime |
| SQLite | Bundled with Python; no separate install |
| Disk space | Local-only storage; sized by memory volume |
CorpusIQ Use Cases¶
| Use Case | How |
|---|---|
| Structured memory upgrade | Replace flat MEMORY.md with queryable vector + FTS5 recall across agent sessions |
| Episodic consolidation | Auto-archive conversation turns; run sleep cycles to consolidate patterns |
| Memory analytics | stats and graph surface what the agent actually remembers and uses |
| Portable backups | export/import for agent migrations between machines |
| Local-first compliance | Zero-cloud memory for sensitive business data |
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.
Limitations / Verification¶
- Provider replaces MEMORY.md/USER.md - migrate existing memory files before enabling
- 4 installs on skills.sh; the repo (2.3K⭐) is established but the Hermes provider integration is new
- Verify:
hermes mnemosyne versionreturns the provider version, and a fresh session shows the# Mnemosyne Memorycontext block
Related¶
- Discovery Page - Aug 12 OpenClaw Ecosystem Sweep
- Knowledge Architecture
- Skills Catalog
- Hermes Changelog -- Latest agent releases and features
- Skills Marketplace -- 946+ community skills
Powered by CorpusIQ