Shared, permissioned memory for coding agents, with an audit row per read
Emerged finds new products worth a look, daily — what’s new today
If you want a household or team assistant without everyone sharing one memory, this gives each signed-in identity its own graph database, enforced at the server rather than by prompt.
A Go agent platform that installs as a Docker Compose appliance, locally or onto a Linux box over SSH, bringing its own Postgres, graph store and object storage. Each person signs in to their own identity and gets an isolated database, where facts are kept as a graph with sources and validity windows rather than a transcript. The agent loop streams, carries a budget and gates terminal and filesystem access. Scheduled jobs, MCP tools, CLI, Telegram and web channels. You pick the model: a router, a chat plan, bundled llama.cpp or Ollama, and the docs say only the local one keeps context home.
Shared, permissioned memory for coding agents, with an audit row per read
Self-hosted AI coworkers, each with its own computer and a policy gateway
Durable AI agents you can self-host, on any model and any machine
A self-hosted index of your email, chats, health and files for any AI agent
Self-hosted company OS that runs Claude Code and Codex agents as staff
A credential vault and approval gate between AI assistants and your apps