agentmem v0.1.0

agentmem v0.1.0

The first public release of agentmem: a lightweight, vectorless memory layer for AI coding agents. Memories stay in plain Markdown files on your machine, while the memory binary gives agents and humans a consistent way to find and manage them.

Highlights

• Local-first Markdown storage with no embeddings, database, cloud account, or background daemon
• A focused CLI for adding, listing, reading, searching, editing, removing, and reindexing memories
• Nested memory names and configurable storage through AGENTMEM_DIR
• Native setup for Claude Code and OpenCode, plus a portable skill installer for other shell-capable agents
• An MCP stdio server for MCP-first or sandboxed environments
• A local loopback web UI for browsing and editing the memory store

Install

curl -fsSL https://agentmem.thamtech.co/install | sh

You can also download the archive for your platform below. Archives contain the memory executable; SHA256SUMS contains checksums for every release asset.

Supported platforms

• Linux: x86-64 and ARM64
• macOS: Intel and Apple Silicon
• Windows: x86-64

Get started

memory add project-notes --summary "Decisions and conventions"
memory list
memory search "conventions"

# Integrate with an agent harness
memory install claude-code
memory install opencode
memory install skill

Run memory ui for the local web interface or memory serve to start the MCP server over stdio.

Release verification

The published binaries were built from agentmem-core commit 54e69aa2a786485a5e7e3916b1ee7e9971d926b0. The release pipeline smoke-tested Linux x86-64, macOS ARM64, and Windows x86-64 artifacts before publishing them.

agentmem is free to use for personal and commercial projects. The source repository remains private for now.