Project management
Create, open, rename, and delete local projects, each with its own file tree.
An AI-native project manager built on Sui, with persistent on-chain memory for Claude, Cursor, VS Code, and any MCP-compatible agent.
Beluga connects AI agents directly to projects on your machine, then gives those agents persistent memory through Walrus Memory.
Decisions, facts, and context are stored as encrypted, vectorized memory on the Walrus network, so a new chat session or a different model can pick up where the last one stopped.
Create, open, rename, and delete local projects, each with its own file tree.
Any MCP-compatible client can connect to Beluga and read or write files inside your projects.
Context survives ended chats, model changes, and machine switches through encrypted Walrus Memory.
Attach one or several Walrus Memory accounts or namespaces to the same project.
Use an integrated wallet to create Walrus Memory accounts now, with agent signatures and on-chain actions planned.
Designed for Sui dApps and smart contracts, but able to manage any kind of project.
Beluga turns the documentation flow into a short local setup path: install the app, connect Sui, create memory, attach it to work, and let an MCP client recall the project state.
Use the macOS or Windows download when available, or build from source.
Import or generate a Sui wallet, with a small amount of SUI for the on-chain memory transaction.
Create or import an encrypted memory account, optionally using a namespace such as default to separate context.
Beluga creates the project, links memory, and scaffolds starter files including WALRUS.md, CLAUDE.md, and README.md.
Point Claude Desktop, Cursor, or VS Code at Beluga and ask it to open the project and check where you left off.
Use a scoped delegate key from Walrus Playground for memory operations instead of exposing a main wallet key to an agent.
Capabilities
Agents store and retrieve project context as encrypted, vectorized memory on the Walrus network.
Beluga ships with an integrated Sui wallet for creating Walrus Memory accounts and future on-chain agent interactions.
Claude Desktop, Cursor, VS Code, and other MCP-compatible clients connect to Beluga over a local MCP endpoint.
Use scoped delegate keys for memory operations instead of exposing a main wallet key to an agent.
Memory lives on Walrus and is keyed to your account and project, not to a single chat window.
Generate or import a Sui-compatible wallet. Keep a small amount of SUI for creating Walrus Memory accounts.
Create or import an on-chain encrypted memory store, with an optional namespace such as default.
Create a project, attach one or more memory accounts, and let Beluga scaffold starter files.
Point Claude, Cursor, or VS Code at Beluga. Agents call recall() at session start and remember() when context matters.
Beluga exposes project and memory operations through the Model Context Protocol. Claude Desktop, Cursor, and VS Code are supported now; GPT-4 client support is listed as coming soon.
Add this server block to claude_desktop_config.json or .cursor/mcp.json at the project root.
{
"mcpServers": {
"walrus-memory": {
"command": "npx",
"args": [
"mcp-remote",
"http://0.0.0.0:47823/mcp",
"--allow-http"
]
}
}
}project_listList projects with path, file count, and last modified time.project_openOpen a project, file tree, and attached memory credentials.project_createCreate a project with starter files.project_renameRename an existing workspace.project_deleteRemove a project when it is no longer needed.file_readRead a file inside the active project.file_writeWrite changes and optionally trigger recall plus remember.file_renameMove or rename a project file.file_deleteDelete a file from the project.folder_createCreate, delete, or rename project folders.recallSearch prior project context at the start of a session.rememberSave important context as a vector embedding.analyzeExtract discrete facts from a longer passage and store them.get_healthCheck relayer connectivity and memory account status.get_account_infoReturn wallet address, account ID, and network details.Everything a builder needs after the landing page: source install commands, Walrus Memory, Sui network inspection, and the MCP protocol spec.
git clone git clone https://github.com/TheRealMagyar/beluga.git
cd beluga
npm install
npm run makeThe installer is generated into the out directory.