Echo is a local MCP server that gives every AI agent you use a shared, permissioned memory space, allowing them to retrieve and compose context from addressable “memory atoms” rather than dumping all stored data. It integrates with Claude Code, Codex, Cursor, and ChatGPT via a single one‑command setup and secures all memories with AES‑256‑GCM encryption, keeping data fast to recall and unreadable at rest.
Funding
Funding not disclosed
Founders
Product
Problem
AI agents often operate in isolation, lacking a shared, organized memory of past interactions, which leads to redundant processing and difficulty retrieving relevant context for new tasks.
Solution
Echo provides a local Memory‑Centered Platform (MCP) that creates a shared, permissioned memory store for multiple AI agents. By installing a single npm package, users connect agents such as Claude Code, Codex, Cursor, and ChatGPT to the Echo server with one command. The platform stores each piece of information as an addressable “memory atom” and links them through relationships, enabling agents to retrieve only the most relevant notes instead of dumping all data. Encryption with AES‑256‑GCM secures memories at rest and in transit, while the optional Vault feature ensures that only the user’s password can unlock the data. Echo’s design is portable and platform‑agnostic, allowing the memory store to follow the user across tools without vendor lock‑in.
Target Audience
Echo is aimed at individual developers, data scientists, and power users who employ multiple AI coding assistants and want a unified, secure local memory layer to enhance agent productivity.
Features
- One‑command installation (`npx @echomem/mcp install --all`) that auto‑connects supported agents
- Atomic, addressable memory objects linked by relationships for context‑driven retrieval
- AES‑256‑GCM encryption at rest and in transit, with user‑derived keys for zero‑knowledge security
- Cross‑tool synchronization enabling the same memory store to be accessed from CLI, Chrome extension, and local HUD
- Permissioned access model that lets users control which agents can read or write specific memories
- Fast recall performance with millisecond‑scale decryption and indexing