memex-hermes is a local memory store for AI and Telegram chats. It presents one file on the user’s laptop that can be read by agents such as Claude, Cursor, and OpenClaw, so conversation history from different AI tools can be kept in one place.
The product is built around importing Telegram Desktop exports in HTML or JSON. A daemon watches the Downloads folder for those exports, detects new chats, and prompts for review. The page also describes per-chat consent, the ability to skip selected chats such as therapist or bank conversations, and block patterns for whole categories. It says the data is stored in local SQLite and never leaves the machine. The same page also mentions searching past chats and using memex_search to find specific conversations or extract structured summaries such as action items, blocked work, and people waiting for a reply.
Installation is offered as a one-line terminal command, through an AI agent setup prompt, or manually. The page says the install script auto-fixes npm permissions and installs a daemon plus an auto-context hook, and it describes the process as idempotent. It also states MIT licensing, no cloud, no API keys, and that it survives provider bans by design. The text says the system can bridge context between AI tools, including reading one agent’s history from another through MCP, and it mentions saving research links such as Perplexity shares with tags for later retrieval.
The tool is aimed at people who use multiple AI agents and Telegram and want shared access to prior context without re-explaining it. It is presented as a local, privacy-first memory layer rather than a cloud service.
In the Frameworks & SDKs space, memex-hermes takes a focused approach. It focuses on integrating Hermes Agent with a unified memory system for seamless, searchable agent history across platforms. It is built as an open-source project for AI agent developers. The project is open source (MIT). It ships for the web and the command line, and it can be self-hosted.
It is developed by parallelclaw, and it first shipped in 2026. Development happens publicly on GitHub with 14 stars and 223 commits in the last 90 days. Key capabilities include hermes integration, unified memory, and searchable history. It exposes integrations via an MCP server.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do