Mnemosyne + OpenClaw

Use Mnemosyne as the memory backend for your OpenClaw agents. Two integration paths:

Path 1: Native Memory Provider (recommended)

Install the OpenClaw extra:

pip install "mnemosyne-memory[openclaw]"

Add to your OpenClaw config.yaml:

memory:
  provider: mnemosyne.integrations.openclaw:create_provider
  config:
    db_path: /path/to/memory/data
    bank: my-agent
    enable_triples: true

That's it. OpenClaw discovers and loads the provider automatically.

Path 2: MCP Server (any MCP-compatible client)

If your OpenClaw version uses MCP for tools, use the standard Mnemosyne MCP server:

1. Install MCP support:

pip install "mnemosyne-memory[mcp]"

2. Add MCP config to your OpenClaw setup:

mnemosyne mcp --transport sse --port 8080

3. Configure OpenClaw to connect to the MCP endpoint.

Provider API

The Mnemosyne provider exposes these operations:

Method	Description
store(key, content, metadata)	Store a memory
retrieve(key)	Get memory by key
search(query, limit)	Semantic search
query(query, params)	Filtered query with date/source/topic
delete(key)	Remove memory by key
get_stats()	Provider statistics
health()	Health check

Filtered Queries

provider = MnemosyneProvider(config={"bank": "my-agent"})

# Date range
results = provider.query("project discussion",
    params={"from_date": "2026-01-01", "to_date": "2026-06-01"})

# By source
results = provider.query("bugs",
    params={"source": "code-review"})

# Triple query (if enable_triples=true)
results = provider.query("triple:prefers")

Configuration

Config Key	Default	Description
db_path	~/.hermes/mnemosyne/data/	Path to memory data directory
bank	openclaw	Memory bank name for isolation
enable_triples	false	Enable TripleStore for graph queries