hyperterminal/docs

Quickstart

Set up the Hyperterminal MCP server in 30 seconds

Setup

Get your key and connect a client.

Verify

Run a first prompt to confirm MCP access.

HTTP Transport

Use hosted Streamable HTTP when stdio is not viable.

Technical Details

Environment variables and package-level details.

Get an API Key

You need a free API key. Sign up at hyperterminal.ai/developers to generate one instantly. See Reference for details on key format and limits.

Add the MCP Server

Select your client below to see the specific configuration file or command needed to add Hyperterminal.

In the config examples below, highlighted lines are the values you must set per environment (usually HYPERTERMINAL_API_KEY).

Common config files across popular clients:

Terminal
claude mcp add hyperterminal --env HYPERTERMINAL_API_KEY=ht_live_your_key -- npx -y @hyperterminal-ai/mcp-server

Scope Options: Add these flags before the server name to change where it is installed:

  • --scope user: Makes it available across all your projects.
  • --scope project: Shares it with your team via .mcp.json.
  • Default is local (current project, private to you).

Management Commands:

Terminal
claude mcp list                  # List all servers
claude mcp get hyperterminal     # Get details
claude mcp remove hyperterminal  # Remove

Use /mcp inside a Claude Code session to check server status.

Add the following to your claude_desktop_config.json file.

Configuration Paths:

PlatformPath
macOS~/Library/Application Support/Claude/claude_desktop_config.json
Windows%APPDATA%\Claude\claude_desktop_config.json
Linux~/.config/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "hyperterminal": {
      "command": "npx",
      "args": ["-y", "@hyperterminal-ai/mcp-server"],
      "env": {
        "HYPERTERMINAL_API_KEY": "ht_live_your_key_here"
      }
    }
  }
}

Warning

You must fully quit and relaunch Claude Desktop after editing the config. Changes are only read at startup.

Add the configuration below to .cursor/mcp.json (for the current project) or ~/.cursor/mcp.json (for global availability).

Alternatively, you can use the UI by going to Settings > MCP > Add new MCP server.

mcp.json
{
  "mcpServers": {
    "hyperterminal": {
      "command": "npx",
      "args": ["-y", "@hyperterminal-ai/mcp-server"],
      "env": {
        "HYPERTERMINAL_API_KEY": "ht_live_your_key_here"
      }
    }
  }
}

MCP tools are available in Cursor's Agent mode, not regular chat. You can use ${env:HYPERTERMINAL_API_KEY} in the JSON file to reference system environment variables instead of hardcoding your key.

Add the following to your mcp_config.json file. You can also access this via the UI: Cmd/Ctrl + , > Windsurf Settings > Plugins (MCP servers) > View raw config.

Configuration Paths:

PlatformPath
macOS~/.codeium/windsurf/mcp_config.json
Windows%USERPROFILE%\.codeium\windsurf\mcp_config.json
Linux~/.codeium/windsurf/mcp_config.json
mcp_config.json
{
  "mcpServers": {
    "hyperterminal": {
      "command": "npx",
      "args": ["-y", "@hyperterminal-ai/mcp-server"],
      "env": {
        "HYPERTERMINAL_API_KEY": "ht_live_your_key_here"
      }
    }
  }
}

Use ${env:HYPERTERMINAL_API_KEY} to reference system environment variables. Make sure to click Refresh in the MCP Servers UI after updating.

Terminal
codex mcp add hyperterminal --env HYPERTERMINAL_API_KEY=ht_live_your_key -- npx -y @hyperterminal-ai/mcp-server

Or manually edit your config file:

~/.codex/config.toml
[mcp_servers.hyperterminal]
command = "npx"
args = ["-y", "@hyperterminal-ai/mcp-server"]

[mcp_servers.hyperterminal.env]
HYPERTERMINAL_API_KEY = "ht_live_your_key_here"

Codex uses TOML, not JSON. Project-level configuration goes in .codex/config.toml at your project root. Use /mcp inside a Codex session to view active servers.

Create an mcpServers configuration file inside your project at .continue/mcpServers/hyperterminal.yaml:

.continue/mcpServers/hyperterminal.yaml
name: Hyperterminal
version: 0.0.1
schema: v1
mcpServers:
  - name: Hyperterminal AI News
    type: stdio
    command: npx
    args:
      - "-y"
      - "@hyperterminal-ai/mcp-server"
    env:
      HYPERTERMINAL_API_KEY: ${{ secrets.HYPERTERMINAL_API_KEY }}

Or drop a JSON config (in the same format as Claude Desktop) into .continue/mcpServers/. Continue auto-detects it.

MCP tools are only available in Agent mode. Secrets use the ${{ secrets.VAR }} syntax.

Open the Cline sidebar in VS Code, go to MCP Servers, and click Configure. This will open your cline_mcp_settings.json file.

cline_mcp_settings.json
{
  "mcpServers": {
    "hyperterminal": {
      "command": "npx",
      "args": ["-y", "@hyperterminal-ai/mcp-server"],
      "env": {
        "HYPERTERMINAL_API_KEY": "ht_live_your_key_here"
      },
      "alwaysAllow": [],
      "disabled": false
    }
  }
}

The alwaysAllow array takes tool names that should skip user confirmation. Leave it empty if you want to manually approve each tool call.

Verify It Works

To make sure everything is connected, ask your agent:

"What small models have the best coding benchmarks right now?"

If it returns structured news results with impact levels, entity tags, and model names, you are good to go!

What can you build?

Here are a few things agents do well once connected:

  • Morning briefing — pull the 5 highest-impact stories from the last 24 hours, filtered by topic or entity
  • Competitive monitoring — ask what a specific company shipped this week, get structured events with source URLs
  • Weekly strategic brief — get the digest: 94 events distilled to key developments with "why it matters" analysis, AGI tracking, and a podcast link
  • Deep research — search a topic, get the top result, follow related stories by similarity score — three calls to build a full picture

See Use Cases for complete end-to-end workflows with real API responses.

HTTP Transport

The default setup above uses stdio transport, where the MCP client spawns the server as a local child process. If you need to connect from web-based agents, cloud-hosted tools, or any client that supports Streamable HTTP, use the hosted endpoint instead — no server to run yourself.

Hosted Endpoint

Endpoint
https://hyperterminal.ai/api/mcp

Authenticate with your API key as a Bearer token in the Authorization header.

Connecting a Client

Any MCP client that supports Streamable HTTP can connect. Point it at the hosted endpoint:

In the HTTP examples below, highlighted lines are the required connection values (url and Authorization header).

claude_desktop_config.json
{
  "mcpServers": {
    "hyperterminal": {
      "type": "streamable-http",
      "url": "https://hyperterminal.ai/api/mcp", 
      "headers": {
        "Authorization": "Bearer ht_live_your_key_here"
      }
    }
  }
}
mcp.json
{
  "mcpServers": {
    "hyperterminal": {
      "type": "streamable-http",
      "url": "https://hyperterminal.ai/api/mcp", 
      "headers": {
        "Authorization": "Bearer ht_live_your_key_here"
      }
    }
  }
}
client.ts
import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
import { Client } from "@modelcontextprotocol/sdk/client/index.js";

const client = new Client({ name: "my-app", version: "1.0.0" });
const transport = new StreamableHTTPClientTransport(
  new URL("https://hyperterminal.ai/api/mcp"), 
  {
    requestInit: {
      headers: {
        Authorization: "Bearer ht_live_your_key_here", 
      },
    },
  }
);
await client.connect(transport);

Which Transport Should I Use?

Use stdio if you're connecting from a local IDE or CLI tool. This is the default and the simplest option — the client launches the server for you, no extra setup needed.

Use Streamable HTTP if you're connecting from cloud-hosted agents, web applications, or any environment where you can't spawn a local child process.

stdioStreamable HTTP
SetupOne-liner, zero configPoint at hosted URL + Bearer token
How it worksClient spawns server as a child processClient connects to hyperterminal.ai/api/mcp over HTTPS
Best forClaude Code, Claude Desktop, Cursor, Windsurf, ClineCloud agents, web apps, Postman, custom integrations
Multiple clientsOne client per processMany clients, one endpoint
NetworkLocal only (stdin/stdout)Works over any network

Not sure? Start with stdio. You can always switch to HTTP later — both transports expose the exact same tools and resources.

Technical Details

Environment Variables

VariableRequiredDefaultDescription
HYPERTERMINAL_API_KEYYes-Your API key (starts with ht_live_)
HYPERTERMINAL_BASE_URLNohttps://hyperterminal.aiOverride for testing

Package Info

CategoryDetails
npm Package@hyperterminal-ai/mcp-server
Transportsstdio (npm package), Streamable HTTP (hosted)
LicenseMIT

Next Steps

Tools

Explore the 6 available tools and real parameter examples

Resources

Auto-loaded context that gives your agent ambient awareness

Examples

Discover developer workflows and a prompt cookbook

Hyperterminal MCP

Real-time AI industry knowledge for your agent

Tools

All 6 MCP tools with examples and parameter reference

On this page

Get an API KeyAdd the MCP ServerVerify It WorksWhat can you build?HTTP TransportHosted EndpointConnecting a ClientWhich Transport Should I Use?Technical DetailsEnvironment VariablesPackage InfoNext Steps