# Akash Network MCP > Akash Network MCP connects your AI client directly to decentralized cloud resources. You can create and manage compute deployments using standard SDL manifests, secure GPU capacity through provider bids and leases, and monitor escrow funds—all programmatically. It lets you run high-performance AI workloads without being locked into a single centralized vendor. ## Overview - **Category:** cloud-infrastructure - **Price:** Free - **Tags:** gpu-computing, decentralized-cloud, sdl-manifest, akash-network, infrastructure-as-code ## Description Running demanding applications used to mean dealing with opaque dashboards and complex billing cycles from massive cloud providers. You'd spend time clicking through menus just to provision or check resource status. This MCP changes that. Your agent can now orchestrate the entire lifecycle of your compute cluster using natural language commands. You write a manifest, tell your AI client to deploy it, and watch the resources spin up across a decentralized network of providers. You'll manage everything from monitoring escrow balances for unexpected costs to creating leases when you need guaranteed capacity. Because this MCP is hosted on Vinkius, you connect once through your preferred agent—whether that’s in VS Code or Cursor—and gain immediate access to the entire catalog of APIs needed for advanced infrastructure management. ## Tools ### add_deposit Adds specified USD funds directly to a deployment's escrow account. ### close_deployment Stops and removes an active compute deployment, releasing associated resources. ### create_deployment Launches a new workload by processing a provided Stack Definition Language (SDL) manifest. ### create_lease Accepts provider bids and formally creates a lease to guarantee compute resources for your deployment. ### enable_auto_top_up Configures the deployment settings so that services can automatically replenish funds when needed. ### get_deployment_settings Fetches the current auto-top-up and billing configuration rules for an active deployment. ### get_deployment Retrieves complete status details for a specific, existing workload or deployment. ### get_provider Gets specific technical details about a known cloud provider within the network. ### list_bids Checks the network to see what bids are available from providers for your current deployment needs. ### list_deployments Retrieves a list of all currently active and paused deployments under your account. ### list_providers Lists every available provider node connected to the Akash network. ### update_deployment_settings Modifies the non-core operational settings, such as auto top-up policies or billing limits. ### update_deployment Makes changes directly to an active deployment's core configuration parameters. ## Prompt Examples **Prompt:** ``` List all my active deployments on Akash. ``` **Response:** ``` I've retrieved your deployments. You have 2 active workloads: 'AI-Training-Cluster' (DSEQ: 105542) and 'Web-App-Node' (DSEQ: 105548). Would you like to check the escrow balance for either of them? ``` **Prompt:** ``` Add a $10 deposit to deployment 105542. ``` **Response:** ``` Processing deposit... I have successfully added $10.00 USD to the escrow for DSEQ 105542. The new balance is now $14.50 USD, extending your estimated runtime. ``` **Prompt:** ``` Close the deployment with DSEQ 105548. ``` **Response:** ``` Closing deployment... DSEQ 105548 has been successfully closed. Associated resources have been released and the remaining escrow balance will be returned to your account. ``` ## Capabilities ### Launch Workloads from Manifests Create new, fully defined compute deployments by referencing a standard Stack Definition Language (SDL) manifest. ### Secure Compute Capacity Poll for bids from network providers and finalize the acquisition of resources by creating formal leases. ### Manage Billing and Funding Monitor deployment escrow balances and add USD deposits to ensure continuous service operation. ### Automate Scaling Policies Enable auto top-up settings so your decentralized infrastructure scales automatically as usage increases. ### Audit Deployment Status Retrieve full details on active deployments, leases, and provider information to maintain operational visibility. ## Use Cases ### Scaling a Live AI Model An AI researcher needs their model to run 24/7. They first check the current resource limits using `get_deployment_settings`. When they see usage climbing, they instruct their agent to use `enable_auto_top_up` and then monitor the continuous funding via `add_deposit`. ### Quickly Testing Provider Availability A developer is building a new service. They start by running `list_providers` to see who's online, then use `get_provider` on a specific name to verify its hardware type before drafting the `create_deployment` manifest. ### Shutting Down Test Clusters A Web3 developer finishes testing a temporary dApp. Instead of logging into three different consoles, they tell their agent to run `close_deployment`, which cleans up all associated resources and prevents unexpected charges. ### Mid-Cycle Configuration Changes A team needs to shift resource allocation for a running service. They use the MCP to call `update_deployment_settings` first, ensuring they change the scaling rules before calling `update_deployment` to commit the final changes. ## Benefits - You avoid manual resource management. Instead of jumping between billing and compute dashboards, you can use `get_deployment` or `list_deployments` to check status instantly via your agent. - Resource bidding is simple. You don't need to wait for emails; just run `list_bids` to poll available provider costs and decide when to commit resources with `create_lease`. - Billing anxiety vanishes. By using `get_deployment_settings`, you can check the auto-top-up status, or use `add_deposit` if you know your service's run time is about to expire. - You maintain control over infrastructure changes. Need to adjust how scaling works? Use `update_deployment_settings` instead of touching core deployment logic with `update_deployment`. - It simplifies complex deployments. You can initiate an entire workload using the `create_deployment` tool just by giving it your SDL manifest, bypassing manual CLI setup. ## How It Works The bottom line is that you manage complex infrastructure provisioning through conversation, not clicking through dashboards. 1. Subscribe to this MCP in Vinkius and provide your Akash API Key. 2. Direct your AI client (like Cursor or Claude) to execute a deployment task, providing the necessary SDL manifest details. 3. Your agent interacts with the network endpoints, creating resources, securing leases, and updating status via simple chat prompts. ## Frequently Asked Questions **How do I keep my deployment funded without manually topping up? (add_deposit)** You use the `enable_auto_top_up` tool to set the policy. This allows your deployed service to automatically replenish its funds when the escrow balance dips below a defined threshold. **What is the best way to check available GPU power? (list_bids)** You run `list_bids` on your deployment. This polls the network and shows you real-time pricing from multiple providers, letting you compare costs before committing to a lease. **Can I update my cluster settings after it's running? (update_deployment_settings)** Yep. Use `update_deployment_settings` for changes like adjusting the auto-top-up limits or changing billing rules without altering the core definition of your workload. **I finished testing, how do I stop and clean up my resources? (close_deployment)** Always use `close_deployment`. This safely terminates the active workload, releases all associated provider resources, and ensures any remaining escrow balance is returned to your account. **What does `list_providers` show me about network availability?** It returns a list of all available compute providers on the Akash Network. You can use this to compare specs, check geographic locations, and assess current capacity before you decide where to deploy your workload. **What information must I provide when I use `create_deployment`?** You need a full Stack Definition Language (SDL) manifest. This file is essential because it defines the entire resource stack, including replicas and hardware requirements for your application. **If I run `get_deployment`, what details can I retrieve about a live workload?** You get comprehensive data on the deployment's current status, total resource usage, and active lease information. This is critical for monitoring high availability in production. **How do I change an active workload using `update_deployment`?** Simply specify the existing deployment identifier along with the new settings you want to apply. The MCP then manages updating those resources without needing a full recreation of the stack. **How do I check for available provider bids after creating a deployment?** Use the `list_bids` tool with your deployment's DSEQ. It typically takes 30-60 seconds for providers to submit bids for your workload. **Can I update a running deployment with a new SDL manifest?** Yes! Use the `update_deployment` tool. Provide the existing DSEQ and your revised SDL string to apply changes to your active resources. **How do I prevent my deployment from closing due to insufficient funds?** You can use `add_deposit` to manually add USD to the escrow, or use `enable_auto_top_up` to configure automatic funding based on your deployment's needs.