Integrates with Hermes (an OpenClaw agent) to provide persistent memory, A2A sharing, and reasoning capabilities, configured via direct URL or local stdio server.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@LogicMem MCP Server Remember that I prefer Telegram for urgent messages."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
This repository is no longer maintained. The canonical source for the LogicMem Python SDK and MCP connector is https://github.com/LogicMem/LogicMem-mcp-. Please file issues and pull requests there.
LogicMem — Prove what your AI did. Tamper-evident records of AI agent actions and memory, built for teams in regulated fintech and insurance.
Website: https://logicmem.io
Package: pip install logicmem (https://pypi.org/project/logicmem/)
LogicMem is designed to support record-keeping duties such as EU AI Act Article 12; it is not a compliance certificate, and nothing here is legal advice.
License: MIT (see LICENSE).
Send a heartbeat to the A2A registry.
Inputs: agent_id (required), client_id (required), status.
List all registered agents for a client_id.
Inputs: client_id (required).
Register this agent in the A2A registry.
Inputs: agent_id (required), client_id (required), name, agent_type.
Check for new shared memories from other agents since last sync.
Inputs: agent_id (required), client_id (required), limit, since_timestamp.
Check if the LogicMem memory server is healthy.
Inputs: no named parameters.
Store a new memory. The agent's memory is permanent and searchable across sessions. Be specific — include who, what, when, why.
Inputs: text (required), tags, source, category, +2 more.
Record whether a stored memory was useful. Feeds the DPO training pipeline.
Inputs: success (required), memory_ids (required), magnitude.
Search memory for relevant entries.
Inputs: query (required), limit, client_id.
Self-critique: evaluate a draft answer against retrieved memories.
Inputs: question (required), draft_answer (required), memory_query.
Full memory session briefing — recall + sentiment + constraints + intelligence + gaps.
Inputs: client_id.
Get memory system stats — total entries, categories, storage.
Inputs: no named parameters.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Scored across 12 tools
Each tool targets a distinct operation: memory log, recall, session, stats, health, outcome, and reflect are all clearly differentiated. The A2A tools (register, heartbeat, list, write_shared, sync) also have distinct purposes with no overlap.
All tools follow a consistent 'logicmem_<domain>_<action>' pattern using snake_case. Minor inconsistency arises because the final token is sometimes a noun (session, stats, health) and sometimes a verb (log, recall, reflect), but the overall structure is predictable and readable.
12 tools is well within the ideal 3-15 range. The set covers two clear domains (memory management and A2A communication) with each tool serving a specific function, making the count feel appropriate and not bloated.
The memory lifecycle is well covered with store, search, full session briefing, stats, health, outcome feedback, and reflection. Minor gaps exist like missing memory update/delete and A2A deregistration, but these are not critical for the core workflows.
Persistent, inspectable memory for AI agents with lineage, correction, and a hosted MCP endpoint.
Shared, permission-aware company context for AI agents, with provenance, approvals and audit.
Context and control for any AI. Agent identity, scoped access, and proof across tools.
Persistent memory, knowledge graph, task management and more for AI agents via MCP.
