Skip to content
modelread

Loading devices…

Remembered on this device. Multi-GPU: search “2x”, “8x”, “node”.

Connect your agent

modelread is a remote MCP server (streamable HTTP, no key). Once added, your agent can answer “will this run on my box?”, pick the quant, and write the llama.cpp / vLLM / MLX launch line itself, so model downloads, server configs and evals start from numbers instead of guesses. Tools are read-only and deterministic: your agent does the reasoning on its own tokens.

Server URLhttps://modelread.com/mcp

Listed in the official MCP Registry as com.modelread/modelread.

Claude Code

Terminal, user scope (all projects)
claude mcp add --transport http --scope user modelread https://modelread.com/mcp

Hugging Face (HF MCP settings)

Add the modelread-mcp Space at huggingface.co/settings/mcp, or use its endpoint
https://patrickbdevaney-modelread-mcp.hf.space/gradio_api/mcp/

Claude.ai / Claude Desktop

Settings → Connectors → Add custom connector, then paste the URL
https://modelread.com/mcp

Cursor

One click, or ~/.cursor/mcp.jsonAdd to Cursor
{
  "mcpServers": {
    "modelread": {
      "url": "https://modelread.com/mcp"
    }
  }
}

VS Code (Copilot agent mode)

One click, or the command lineAdd to VS Code
code --add-mcp '{"name":"modelread","type":"http","url":"https://modelread.com/mcp"}'

OpenAI Codex CLI

Terminal, or ~/.codex/config.toml
codex mcp add modelread --url https://modelread.com/mcp

# or in ~/.codex/config.toml
[mcp_servers.modelread]
url = "https://modelread.com/mcp"

Gemini CLI

Terminal, or settings.json (httpUrl)
gemini mcp add --transport http modelread https://modelread.com/mcp

Windsurf

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "modelread": {
      "serverUrl": "https://modelread.com/mcp"
    }
  }
}

Zed

settings.json
{
  "context_servers": {
    "modelread": {
      "url": "https://modelread.com/mcp"
    }
  }
}

LM Studio (0.3.17+)

One click, or ~/.lmstudio/mcp.json: your local model can then size other modelsAdd to LM Studio
{
  "mcpServers": {
    "modelread": {
      "url": "https://modelread.com/mcp"
    }
  }
}

Tools

estimate_fit
repo + device + context → verdict, best quant, tok/s range, max context, memory placement, launch command
compare_hardware
one repo across up to 8 devices, side by side
find_models
"best model for task X on device Y" with license, speed and size filters
search_models
search the Hub index by name or task
list_hardware
the device catalog (filterable): GPUs, Macs, phones, nodes, cloud instances with $/h
ask
free-text question → the same ranked answer the website gives

Hardware accepts preset ids, “2x-rtx3090”, free text (“MacBook Pro M3 Max 128GB”, “AWS p5.48xlarge”, “Galaxy S25 Ultra”) or a custom spec like gpu=24,ram=64,bw=936.

Agent skill

A drop-in SKILL.md teaches Claude Code (and any agent that reads skills) when to call modelread and how to read its ranges before downloading or serving a model.

mkdir -p ~/.claude/skills/modelread && curl -fsSL https://modelread.com/skill.md -o ~/.claude/skills/modelread/SKILL.md

HTTP API

Same answers as JSON. Rate-limited per client; schema at /openapi.json, summary for LLMs at /llms.txt.

curl "https://modelread.com/api/estimate?repo=Qwen/Qwen3.6-27B&hw=rtx4090&ctx=32768"

Fit, speed, memory placement and launch line for one repo on one device.

curl -N "https://modelread.com/api/ask/stream" -H 'content-type: application/json' \
  -d '{"q":"best coding model for two RTX 3090s with 128GB RAM"}'

Server-sent events: status → query → draft → final.

curl "https://modelread.com/api/hardware"

Every device preset with memory, bandwidth, compute, backend and price.

Askwhat should I run?
Browsebest models per device
Devicesall hardware
Connect your agentMCP · API
How the numbers workcalibration