# Install Hugging Bay in your agent

Pick your agent, copy and paste, and you are set. No account or key.

These setups use the four core tools: search models, check what fits your computer, read one model's files and hashes, and get a plain answer on whether to serve a file. All four only read. For board, request and plan tools, change profile=core to profile=discovery.

## Claude Code

Status: Connected in our test on 2026-11-05.

1. Paste the command in your terminal. It adds Hugging Bay for all your projects.
2. Optional: add the skill so Claude knows when to use it.
3. Run /mcp in Claude Code to see hugging-bay connected.

Add the server:

```
claude mcp add --transport http --scope user hugging-bay 'https://huggingbay.xyz/api/mcp?profile=core'
```

Add the skill (optional, macOS and Linux):

```
d="$HOME/.claude/skills/hugging-bay-models"; mkdir -p "$d/scripts" && curl -fsS https://huggingbay.xyz/skills/hugging-bay-models/SKILL.md -o "$d/SKILL.md" && curl -fsS https://huggingbay.xyz/skills/hugging-bay-models/scripts/hb.py -o "$d/scripts/hb.py"
```

Guide: https://code.claude.com/docs/en/mcp

## Claude Desktop

Status: Not tested by us yet.

1. Open Settings, then Connectors, and choose Add custom connector.
2. Name it hugging-bay, paste the URL, and leave sign-in off.
3. On a plan without custom connectors, use the config file instead and restart Claude. It needs Node.js 22 or later.

Connector URL:

```
https://huggingbay.xyz/api/mcp?profile=core
```

Config file on macOS: ~/Library/Application Support/Claude/claude_desktop_config.json:

```
{
  "mcpServers": {
    "hugging-bay": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://huggingbay.xyz/api/mcp?profile=core"
      ]
    }
  }
}
```

Config file on Windows: %APPDATA%\Claude\claude_desktop_config.json:

```
{
  "mcpServers": {
    "hugging-bay": {
      "command": "cmd",
      "args": [
        "/c",
        "npx",
        "-y",
        "mcp-remote",
        "https://huggingbay.xyz/api/mcp?profile=core"
      ]
    }
  }
}
```

Guide: https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp

## Codex CLI

Status: Config accepted, not connected in our test.

1. Paste the command, or add the block to ~/.codex/config.toml.
2. Start Codex and run /mcp to see hugging-bay.

Add the server:

```
codex mcp add hugging-bay --url 'https://huggingbay.xyz/api/mcp?profile=core'
```

Or ~/.codex/config.toml:

```
[mcp_servers.hugging-bay]
url = "https://huggingbay.xyz/api/mcp?profile=core"
```

Add the skill (optional, macOS and Linux):

```
d="$HOME/.codex/skills/hugging-bay-models"; mkdir -p "$d/scripts" && curl -fsS https://huggingbay.xyz/skills/hugging-bay-models/SKILL.md -o "$d/SKILL.md" && curl -fsS https://huggingbay.xyz/skills/hugging-bay-models/scripts/hb.py -o "$d/scripts/hb.py"
```

Guide: https://developers.openai.com/codex/mcp

## Cursor

Status: Not tested by us yet.

1. Open the install link and confirm in Cursor.
2. Or merge the JSON into ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project).
3. Check Settings, Tools & MCP: hugging-bay should be on.

Install link: cursor://anysphere.cursor-deeplink/mcp/install?name=hugging-bay&config=eyJ1cmwiOiJodHRwczovL2h1Z2dpbmdiYXkueHl6L2FwaS9tY3A%2FcHJvZmlsZT1jb3JlIn0%3D

mcp.json:

```
{
  "mcpServers": {
    "hugging-bay": {
      "url": "https://huggingbay.xyz/api/mcp?profile=core"
    }
  }
}
```

Guide: https://cursor.com/docs/context/mcp

## Gemini CLI

Status: Connected in our test on 2026-11-05.

1. Paste the command. It adds Hugging Bay for all your projects.
2. Gemini turns MCP off in folders you have not trusted. Trust your project folder when it asks.
3. Run /mcp in Gemini to see hugging-bay connected.

Add the server:

```
gemini mcp add --scope user --transport http hugging-bay 'https://huggingbay.xyz/api/mcp?profile=core'
```

Or ~/.gemini/settings.json:

```
{
  "mcpServers": {
    "hugging-bay": {
      "url": "https://huggingbay.xyz/api/mcp?profile=core",
      "type": "http"
    }
  }
}
```

Guide: https://geminicli.com/docs/tools/mcp-server/

## VS Code

Status: Not tested by us yet.

1. Open the install link, or run the command with the code CLI.
2. Or merge the JSON into .vscode/mcp.json in your project.
3. Run MCP: List Servers, start hugging-bay, and use Agent mode in Copilot Chat.

Install link: vscode:mcp/install?%7B%22name%22%3A%22hugging-bay%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fhuggingbay.xyz%2Fapi%2Fmcp%3Fprofile%3Dcore%22%7D

Add with the code CLI:

```
code --add-mcp '{"name":"hugging-bay","type":"http","url":"https://huggingbay.xyz/api/mcp?profile=core"}'
```

.vscode/mcp.json:

```
{
  "servers": {
    "hugging-bay": {
      "type": "http",
      "url": "https://huggingbay.xyz/api/mcp?profile=core"
    }
  }
}
```

Guide: https://code.visualstudio.com/docs/copilot/customization/mcp-servers

## Windsurf

Status: Not tested by us yet.

1. Merge the JSON into ~/.codeium/windsurf/mcp_config.json.
2. Refresh MCP servers in Cascade settings, or restart Windsurf.

mcp_config.json:

```
{
  "mcpServers": {
    "hugging-bay": {
      "serverUrl": "https://huggingbay.xyz/api/mcp?profile=core"
    }
  }
}
```

Guide: https://docs.windsurf.com/windsurf/cascade/mcp

## Hermes Agent

Status: Connected in our test on 2026-11-05.

1. Add the server to ~/.hermes/config.yaml.
2. Add the skill so Hermes knows when to use it.
3. Run hermes mcp test hugging-bay, then restart Hermes. If it says HTTP is not available, reinstall with pip install 'hermes-agent[mcp]'.

~/.hermes/config.yaml:

```
mcp_servers:
  hugging-bay:
    url: "https://huggingbay.xyz/api/mcp?profile=core"
```

Add the skill (macOS and Linux):

```
d="$HOME/.hermes/skills/hugging-bay-models"; mkdir -p "$d/scripts" && curl -fsS https://huggingbay.xyz/skills/hugging-bay-models/SKILL.md -o "$d/SKILL.md" && curl -fsS https://huggingbay.xyz/skills/hugging-bay-models/scripts/hb.py -o "$d/scripts/hb.py"
```

Guide: https://hermes-agent.nousresearch.com/docs/user-guide/features/mcp

## OpenClaw

Status: Connected in our test on 2026-11-05.

1. Paste the command. OpenClaw connects once before it saves.
2. Run openclaw mcp probe to list the tools.

Add the server:

```
openclaw mcp add hugging-bay --url 'https://huggingbay.xyz/api/mcp?profile=core' --transport streamable-http
```

Guide: https://docs.openclaw.ai/cli/mcp

## LangChain

Status: Connected in our test on 2026-11-05.

1. Install the adapter, then run the example. It prints the tool names.

Install:

```
pip install langchain-mcp-adapters
```

Python:

```
import asyncio
from langchain_mcp_adapters.client import MultiServerMCPClient

async def main():
    client = MultiServerMCPClient({
        "hugging-bay": {"transport": "streamable_http", "url": "https://huggingbay.xyz/api/mcp?profile=core"},
    })
    tools = await client.get_tools()
    print([tool.name for tool in tools])
    # Give the tools to your agent, for example create_agent(model, tools).

asyncio.run(main())
```

Guide: https://docs.langchain.com/oss/python/langchain/mcp

## LlamaIndex

Status: Connected in our test on 2026-11-05.

1. Install the MCP tools package, then run the example. It prints the tool names.

Install:

```
pip install llama-index-tools-mcp
```

Python:

```
import asyncio
from llama_index.tools.mcp import BasicMCPClient, McpToolSpec

async def main():
    spec = McpToolSpec(client=BasicMCPClient("https://huggingbay.xyz/api/mcp?profile=core"))
    tools = await spec.to_tool_list_async()
    print([tool.metadata.name for tool in tools])
    # Give the tools to your agent, for example FunctionAgent(tools=tools, llm=llm).

asyncio.run(main())
```

Guide: https://developers.llamaindex.ai/python/framework/module_guides/deploying/agents/tools/#mcp-tools

## OpenAI Agents SDK

Status: Connected in our test on 2026-11-05.

1. Install the SDK, then run the example. It prints the tool names; uncomment the last line to ask a question.

Install:

```
pip install openai-agents
```

Python:

```
import asyncio
from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp

async def main():
    async with MCPServerStreamableHttp(name="hugging-bay", params={"url": "https://huggingbay.xyz/api/mcp?profile=core"}, cache_tools_list=True) as hugging_bay:
        print([tool.name for tool in await hugging_bay.list_tools()])
        agent = Agent(name="Assistant", instructions="Use Hugging Bay to find open models.", mcp_servers=[hugging_bay])
        # result = await Runner.run(agent, "Find a chat model that fits 16 GB of RAM.")

asyncio.run(main())
```

Guide: https://openai.github.io/openai-agents-python/mcp/