MCP Server — akash-network/mcp

Akash MCP Server

What you get

This project is still under development and not all tools are available. We are working on adding more tools and improving the server. We are open to contributions! See CONTRIBUTING.md for more details.

This server provides a bridge between AI agents and the Akash Network, allowing AI models to deploy applications, create leases, manage deployments, and interact with other Akash services directly through typed tools. It implements the Model Context Protocol, making it compatible with various AI platforms and tools.

• Wallet and Client Management — Handles Akash wallet authentication and client initialization
• Certificate Management — Manages Akash certificates
• Tools for Akash Interaction —
• Account address retrieval
• Deployment creation, querying, updating, and termination
• SDL (Stack Definition Language) operations

What the assistant can call

Once MCP is connected, these are the calls the assistant has available:

• GetAccountAddrTool — Retrieve your Akash account address
• GetBalancesTool — Get the AKT (uakt) and other balances for a given Akash account address
• GetBidsTool — Get bids for deployments
• CreateDeploymentTool — Create a new deployment on Akash Network
• GetDeploymentTool — Get deployment details including status, groups, and escrow account
• CloseDeploymentTool — Close/terminate a deployment on Akash Network
• AddFundsTool — Deposit additional AKT (uakt) into a deployment escrow account
• GetSDLsTool — Get a list of available SDLs (from awesome-akash repository)
• GetSDLTool — Get a specific SDL by name
• SendManifestTool — Send a manifest to a provider
• CreateLeaseTool — Create a lease with a provider
• GetServicesTool — Get information about active services

Configuration and credentials

• Node.js (v18 or later) - An Akash wallet (a mnemonic or a hex-encoded private key is required for interaction with Akash Network) - Basic knowledge of Akash Network

More MCP options

Compare MCP servers by category, or browse hosted MCP servers if you prefer an endpoint over a local process.