Developer API

Build apps where users are
free to think.

OpenAI-compatible inference that runs locally. Point your existing SDK atlocalhost:11435and everything just works. Nothing stored. Ever.

Quickstart

Python
from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:11435/v1",
    api_key="not-needed"  # local inference
)

response = client.chat.completions.create(
    model="tarx",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True
)

for chunk in response:
    print(chunk.choices[0].delta.content, end="")

Install TARX first: curl -fsSL cli.tarx.com/install | sh

Endpoints

POST/v1/chat/completions
GET/v1/models
POST/v1/embeddings
GET/health

Privacy guarantee

TARX local inference stores zero data beyond your machine. No telemetry on prompts or responses. No training on your data. The Supercomputer mesh uses ephemeral compute — queries are processed and discarded. Set privacy_flag=true to ensure queries never leave your hardware.

Pricing

Local

Free

Run on your hardware

  • + Unlimited inference
  • + Perfect memory
  • + No account needed
  • + Full privacy

Supercomputer

Free tier

Mesh-augmented inference

  • + 100 requests/day free
  • + Distributed compute
  • + Earn credits contributing
  • + Privacy flag available

Enterprise

Contact us

Custom deployment

  • + Dedicated mesh subnet
  • + Custom endpoints
  • + Team management
  • + SLA guarantee

MCP Integration

Add TARX tools to Claude Code, Cursor, or any MCP-compatible client.

shell
tarx mcp add claude      # adds to ~/.claude.json
tarx mcp add cursor      # adds to cursor config
tarx mcp add vscode      # adds to VS Code