What models can I provision?

Llama 3.1, Qwen 2.5, Mistral, Gemma 2, and Phi-3 on dedicated GPUs.

Is billing usage-based?

Yes, Auxen charges strictly per minute of instance runtime with no subscriptions.

auxen

Verified

Provisions private AI model endpoints on dedicated GPUs.

MCP ServerAI & KnowledgeRemote (streamable-http)

View on GitHub

Updated 2026-06-15

What is the auxen MCP server?

Auxen provisions private, dedicated GPU instances running open-source models such as Llama 3.1, Qwen 2.5, Mistral, Gemma 2, and Phi-3. Each instance delivers fully private inference with no shared resources or third-party routing.

Billing is strictly pay-per-minute with no subscriptions required. The server is accessed remotely using streamable-http and supports both OAuth 2.1 + PKCE and direct API key authentication.

Install & connect

Add this to your MCP client config. Pick your client below and copy.

{
  "mcpServers": {
    "auxen": {
      "url": "https://api.auxen.ai/mcp"
    }
  }
}

Tools it exposes

6 tools your AI client can call once connected.

auxen_list_models— List available models, optionally filtered by size

auxen_get_instance_status— Get status, endpoint, api_key for an instance

auxen_list_instances— List all instances on the account

auxen_get_balance— Read USD credits + active subscriptions

auxen_provision_model— Provision a new model instance — spends money

auxen_destroy_instance— Destroy an instance — irreversible

Example prompts

Once connected, try asking your AI client:

“List all available models supported by Auxen

“Show the status of my current model instances

“Provision a new Llama 3.1 instance on a dedicated GPU

“Check my Auxen account balance and credits

Security & permissions

Remote server accessed via streamable-http at api.auxen.ai/mcp. Requires OAuth 2.1 + PKCE or an Auxen API key (auxen_live_* or auxen_test_*) sent as Authorization: Bearer.

What you can do with auxen

Private Model Deployment

Spin up dedicated GPU instances for sensitive or proprietary AI workloads without shared inference.

Cost Monitoring

Check account balance, credits, and active subscriptions before provisioning new instances.

Instance Management

List, monitor status, or destroy model instances directly from an MCP client.

How to use auxen

1Add https://api.auxen.ai/mcp as a remote MCP server in your client.
2Choose OAuth 2.1 + PKCE for browser-based flows or generate an API key at auxen.ai/dashboard/api-keys.
3Send the key as Authorization: Bearer <key> on MCP calls if using direct authentication.
4Use tools such as auxen_list_models to explore options before provisioning.
5Monitor usage with auxen_get_balance to control pay-per-minute costs.

auxen: pros & cons

Pros

+Fully private dedicated GPUs with no shared inference
+Pay-per-minute pricing with no subscriptions
+Supports popular open models including Llama, Qwen, and Mistral
+OAuth and API key auth options for flexible client integration

Cons

–Provisioning and runtime incur real monetary costs
–Requires an Auxen account and either OAuth or API key setup
–Only remote streamable-http access; no local installation option

Did you find this helpful?

Frequently asked questions

Use OAuth 2.1 + PKCE or send an auxen_live_* or auxen_test_* API key as Authorization: Bearer.

User reviews

Verified reviews from the community shape this listing's rating.

Loading reviews…

Promote auxen