Single-binary CLI for Memoria — bi-temporal AI agent memory served over MCP.
The CLI handles install + auth + MCP config in one shot, and deposits a Claude Code skill so every session in a configured project automatically reads Memoria's mental model.
curl -fsSL https://api.memoria.premex.se/install.sh | sh
memoria init <your-api-key>Mint an API key at https://memoria.premex.se/dashboard/brains/<brain-id>/keys. The CLI stores it in your OS keychain (macOS) or a permission-restricted file (Linux), wires up ~/.claude.json, and deposits the memoria skill at ~/.claude/skills/memoria/SKILL.md. Subsequent Claude Code sessions pick it all up automatically.
- The Go source for the
memoriaCLI (cmd/,internal/). - The in-session Claude Code skill (
skill/SKILL.md) embedded into the binary at build time. - The bootstrap Claude Code plugin (
plugin/memoria-setup/) that installs the CLI on demand. - The installer shell script (
install.sh). - User-facing docs (
docs/): CLI reference, mental model, scopes reference, and worked examples.
The Memoria service itself (api, web app, dashboard) lives in a separate private repo. This repo is the user-facing client + docs surface.
- docs/cli.md — full CLI reference: all commands, flags, token resolution,
headersHelper - docs/mental-model.md — what Memoria stores and why, in plain words
- docs/scopes.md — API key scopes and per-endpoint access control
- docs/examples/glyph-cron-on-memoria.md — migrating a daily-maintenance cron from a wiki to Memoria
Apache-2.0. See LICENSE.