Elroy is a scriptable, memory augmented AI personal assistant, accessible from the command line. It features:
- Long-term Memory: Automatic memory recall of past conversations
- Reminders Track context based and timing based reminders
- Simple scripting interface: Script Elroy with minimal configuration overhead
- CLI Tool interface: Quickly review memories Elroy creates for you, or jot quick notes for Elroy to remember.
The fastest way to get started is using the install script:
curl -LsSf https://raw.githubusercontent.com/elroy-bot/elroy/main/scripts/install.sh | shOr install manually with UV:
# Install UV first
curl -LsSf https://astral.sh/uv/install.sh | sh
# Then install Elroy
uv pip install elroyFor detailed installation instructions including Docker and source installation options, see the Installation Guide.
Once installed, run elroy to open the interactive chat interface:
elroyFor source development with automatic restart on code changes, use:
just devThis requires watchexec and runs Elroy from the local source tree with restart-on-change behavior for .py, .yml, and .yaml files.
It uses the elroy console entrypoint rather than python -m elroy.ui.app, which avoids runpy module re-execution warnings from the package import layout.
Elroy opens a terminal UI where you can chat, create memories, and manage reminders. Use the Textual command palette with Ctrl+P to discover available actions.
Elroy creates ~/.elroy/contrib as the default assistant-editable plugin directory. Python files in that directory can define @tool functions for assistant tools and ELROY_PLUGIN = ElroyPlugin(...) for custom sidebar tabs.
For self-improvements that should not modify core Elroy, ask the assistant to use edit_contrib_with_codex. Elroy can run from any source branch; Codex-backed contrib changes run directly in the home contrib repo, which Elroy initializes and manages as git.
Elroy's tools allow it to create and manage memories and reminders. In the background, redundant memories are consolidated.
As reminders or memories become relevant to the conversation, they are recalled into context. A Relevant Context panel makes all information being surfaced to the assistant available to the user.
All commands available to the assistant are also available to the user through the Textual command palette.
For a guide of what tools are available and what they do, see: tools guide.
For a full reference of tools and their schemas, see: tools schema reference
Elroy is designed to be highly customizable, including appearance and memory consolidation parameters.
Configuration can be provided via environment variables (e.g. ELROY_CHAT_MODEL=gpt-5) or a config file at ~/.elroy/elroy.conf.yaml.
For full configuration options, see configuration documentation.
Elroy supports OpenAI, Anthropic, Google (Gemini), and any OpenAI-compatible API.
The model is selected automatically based on which API key is set, or can be configured explicitly:
# Use an Anthropic model
ELROY_CHAT_MODEL=claude-sonnet-4-5-20250929 elroy
# Use a specific OpenAI model
ELROY_CHAT_MODEL=gpt-5 elroy
# Use Gemini
ELROY_CHAT_MODEL=gemini/gemini-2.0-flash elroyCommon model aliases can be used as the chat_model config value:
sonnet: Anthropic's Claude 4.5 Sonnetopus: Anthropic's Claude 4.5 Opushaiku: Anthropic's Claude 3.5 Haikuo1: OpenAI's o1 model
Elroy provides skills for Claude Code that expose memory management as slash commands:
/remember- Create a long-term memory/recall- Search through memories/list-memories- List all memories/remind- Create a reminder/list-reminders- List active reminders/ingest- Ingest documents into memory
Install the Claude Code skills using the Elroy CLI:
elroy install-skillsOr use the just command from the repository:
just install-claude-skillsThis installs skills to ~/.claude/skills/ making them available in all Claude Code sessions.
Important: Restart your Claude Code session after installation to load the new skills.
To uninstall:
elroy install-skills --uninstallFor detailed usage and examples, see claude-skills/README.md.
main comes with backwards compatibility and automatic database migrations.
stable is sync'd with the latest release branch.
experimental is a test branch with upcoming changes. These may contain breaking changes and/or changes that do not come with automatic database migrations.
Distributed under the Apache 2.0 license. See LICENSE for more information.
Bug reports and feature requests are welcome via GitHub
