mnemopay（mnemopay 搜索）

MnemoPay — Agent Memory + Wallet

Give any AI agent persistent memory and a micropayment wallet. MnemoPay unifies cognitive memory (Mnemosyne) and escrow economics (AgentPay) into a single MCP server. The core innovation: payment outcomes reinforce the memories that led to successful decisions.

Setup

Add the MnemoPay MCP server:

openclaw mcp set mnemopay '{"command":"npx","args":["-y","@mnemopay/sdk"],"env":{"MNEMOPAY_AGENT_ID":"openclaw-agent","MNEMOPAY_MODE":"quick"}}'

For production mode (persistent Postgres + Redis storage):

openclaw mcp set mnemopay '{"command":"npx","args":["-y","@mnemopay/sdk"],"env":{"MNEMOPAY_AGENT_ID":"openclaw-agent","MNEMOPAY_MODE":"production","MNEMO_URL":"http://localhost:8100","AGENTPAY_URL":"http://localhost:3100"}}'

Verify tools are available:

openclaw mcp list-tools mnemopay

Tools Reference

Memory Tools

Payment Tools

Status Tools

Workflows

Workflow 1: Remember and Recall

When you learn something important during a conversation:

1. Call mcp__mnemopay__remember with the content, optionally setting importance (0-1) and tags
2. In future sessions, call mcp__mnemopay__recall with a semantic query to retrieve relevant memories
3. Use recalled memories to provide personalized, context-aware responses

Workflow 2: Charge for Value Delivered

When you complete a task that delivers measurable value:

1. Call mcp__mnemopay__balance to check current reputation and max charge limit
2. Deliver the work
3. Call mcp__mnemopay__charge with amount and clear description of value delivered
4. When user confirms satisfaction, call mcp__mnemopay__settle with the transaction ID
5. Settlement automatically reinforces recently-accessed memories (the feedback loop)

Workflow 3: Session Start Protocol

At the beginning of every conversation:

1. Call mcp__mnemopay__recall with a query related to the user's first message (or no query for top memories)
2. Use recalled memories to greet the user with context from previous sessions
3. Call mcp__mnemopay__profile to check agent health

Workflow 4: Memory Maintenance

Periodically (every 10-20 interactions):

1. Call mcp__mnemopay__consolidate to prune stale memories
2. Review top memories with mcp__mnemopay__recall (limit: 20)
3. Call mcp__mnemopay__reinforce on memories that are still actively useful

The Feedback Loop

MnemoPay's core innovation is connecting memory to economics:

Remember → Recall → Act → Charge → Settle
                                      ↓
                              Reinforce memories (+0.05)
                              Boost reputation (+0.01)

Successful payments automatically strengthen the memories that led to good decisions. Over time, the agent gets better at remembering what matters. Failed transactions (refunds) dock reputation, creating accountability.

Environment Variables

Error Handling

• If a tool call fails, the error message is returned as text. Do not retry immediately — diagnose the issue first.
• charge will fail if amount exceeds $500 x current reputation score. Check balance first.
• settle and refund require a valid transaction ID from a prior charge.
• forget returns whether the memory existed. A "not found" response is not an error.

Links

• SDK: https://github.com/t49qnsx7qt-kpanks/mnemopay-sdk
• npm: https://www.npmjs.com/package/@mnemopay/sdk
• Docs: https://github.com/t49qnsx7qt-kpanks/mnemopay-sdk#readme