home / mcp / akave mcp server

Akave MCP Server

Provides access to Akave storage via MCP tools for bucket and object management, including signed URLs and object operations.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "akave-ai-akave-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "akave-mcp-js"
      ],
      "env": {
        "AKAVE_ENDPOINT_URL": "YOUR_ENDPOINT_URL",
        "AKAVE_ACCESS_KEY_ID": "YOUR_ACCESS_KEY_ID",
        "AKAVE_SECRET_ACCESS_KEY": "YOUR_SECRET_ACCESS_KEY"
      }
    }
  }
}

You train AI models to manage and access your Akave storage through a dedicated MCP server. It provides a standardized set of tools to list, read, write, and manage buckets and objects, generate secure access links, and work with both Claude Desktop and local LLMs via Ollama.

How to use

Interact with your Akave storage by calling the available MCP tools through your MCP client. The server exposes a structured set of actions you can invoke from your AI model, such as listing buckets, listing or reading objects, uploading content, updating objects, and generating signed URLs for secure access. You can compose tool calls in your AI workflow to perform tasks like uploading results from an analysis to a bucket, retrieving files for inclusion in a response, or copying data between locations.

How to install

Prerequisites you need before starting: - Node.js 16+ installed on your machine - Access to an Akave account with: - Access Key ID - Secret Access Key - Endpoint URL - If you plan to use local LLMs via Ollama, you also need Go 1.23+ and Ollama installed.

Create your MCP configuration file. Save it as mcp.json in a location you can reference from your workflow. The following configuration defines a local, stdio MCP server that you run with npx and the akave-mcp-js package. Replace the placeholder values with your actual credentials and endpoint.

{
  "mcpServers": {
    "akave": {
      "command": "npx",
      "args": [
        "-y",
        "akave-mcp-js"
      ],
      "env": {
        "AKAVE_ACCESS_KEY_ID": "your_access_key",
        "AKAVE_SECRET_ACCESS_KEY": "your_secret_key",
        "AKAVE_ENDPOINT_URL": "your_endpoint_url"
      }
    }
  }
}

Troubleshooting and tips

If you encounter connection or credential issues, verify that your Akave credentials and endpoint URL are correctly set in the mcp.json file and that your network can reach the endpoint. If a local LLM setup is used, ensure Ollama is running and that the MCPHost binding is correctly wired to the selected model. For debugging, run MCPHost with the --debug flag and review the logs produced by your environment for detailed error information.

Notes on environment and configuration

Environment variables shown in the example are required for the MCP server to authenticate with Akave storage and to know where to connect. Keep these values secure and do not commit them to public repositories.

Available tools

list_buckets

List all storage buckets in your Akave account.

list_objects

List objects within a specified bucket, with optional prefix filtering.

get_object

Read the contents of an object from a bucket.

put_object

Upload or write a new object to a bucket.

get_signed_url

Generate a signed URL to securely access an object.

update_object

Update the contents or metadata of an existing object.

delete_object

Delete an object from a bucket.

copy_object

Copy an object to another location or bucket.

create_bucket

Create a new storage bucket.

delete_bucket

Delete a storage bucket.

get_bucket_location

Get the region or location of a bucket.

list_object_versions

List all versions of objects when versioning is enabled.