Short, direct replies
Your local AI assistant
Memory.Action.Your control.
Anda Bot brings useful context from past conversations into the work ahead. Give it a task, connect your tools, and build on what matters—with models you choose and memory on your machine.
- Open source
- Memory stored locally
- Your choice of models
01 / Memory
Let useful context accumulate.
Your preferences, projects, and important decisions can become lasting context. Anda Brain organizes useful information from conversations and recalls it when relevant.
- 01
Your working preferences
Carry the details that shape how you like things done into future conversations.
- 02
Connected context
Connect people, projects, events, and decisions in a local knowledge graph.
- 03
Room for things to change
Tell Anda when information changes, so new context can become part of its memory.
02 / Action
Bring your assistant into the work.
Ask questions, explore information, or hand over a longer task. Anda can use the tools you connect and keep working toward a defined goal.
Browse and explore
Read web pages, gather information, and interact with sites through browser tools.
Work with files and tools
Read and write local files, run commands, and extend workflows with skills and MCP.
Keep longer tasks moving
Maintain a goal, coordinate subagents, and carry work forward as context grows.
Make room for routines
Schedule one-off or recurring tasks while your Anda runtime is running.
03 / Your control
Your models. Your memory. Your choices.
Keep a lasting assistant without tying its memory to one model account.
- 01
Data on your machine
Keep configuration, conversations, and memory in your local Anda home directory.
- 02
Models you choose
Connect supported providers and change models while retaining your local memory.
- 03
Open by design
Inspect the source, configure your tools, and adapt the assistant to your own workflow.
Local storage does not mean offline processing. Relevant content may be sent to the model providers and services you configure.
Ways to use Anda
An assistant within reach.
Choose the interface that fits how you work. Each connects to the Anda runtime you configure.
Browser
Keep Anda beside the page you are reading. Connect the Chrome or Edge extension to your local runtime.
Set up the extensionTerminal
Chat, select a workspace, and use local tools from the CLI and terminal interface.
Use the terminalMessaging
Connect Telegram, WeChat, Discord, or Lark/Feishu. Access follows your channel configuration.
Connect a channelDesktop workbench
Chat, memory, browser, terminal, and Git in one desktop app, with a tray that keeps Anda running.
About Anda DesktopGet started
Start with something you want to do.
Install the local runtime, connect a model provider, and open your preferred interface. The current installers include the CLI and launcher where supported.
You need a supported model provider and usually an API key. Model and connected-service usage may incur separate charges.
Anda Desktop
Download the installer. It installs Anda Desktop, the anda command line, curated skills, and a tray that keeps Anda running at login.
- Run setup
- Enter model settings
- Pair browser
The tray shows service status, installs updates, restarts the service, opens logs, and copies browser pairing tokens.
Before you start
Do I need to set up a separate memory service?
Ordinary long-term memory uses your model configuration. You do not need to configure the advanced Brain runtime or learning features to start.
Is everything processed on my computer?
Memory and runtime state are stored locally. Model calls, speech services, and connected tools may process content elsewhere, depending on your configuration.
Can I use just the browser extension?
The extension connects to a running Anda runtime; it is not a standalone hosted assistant. Install and configure Anda first, then pair the extension.
Can it work while the app is closed?
Tasks require a running Anda runtime and access to the configured services. Closing a client and stopping the runtime are different actions; work cannot run while the computer is shut down.
Make room for an assistant of your own.
Start a conversation, connect your tools, and let useful context accumulate through the work you do together.