AI Access

Connect OpenAI, ChatGPT, Claude, Cursor, or any MCP-compatible AI to ChartInspect data.

Let an AI query ChartInspect on-chain metrics, prices, indicators, and economic data directly. AI access requires an active Pro subscription. Every integration is read-only and inherits your plan access and API limits.

Use the AI Access tab in the API Playground to fill the examples with your API key automatically.

#Choose an integration

ClientRecommended connectionAuthentication
OpenAI Responses APIRemote MCP toolPro API key
ChatGPT custom appRemote MCP appChartInspect OAuth
Claude.ai or Claude DesktopCustom connectorChartInspect OAuth
Claude Messages APIMCP connectorPro API key
Claude Code, Cursor, other MCP clientsStreamable HTTP MCPPro API key or OAuth
Custom GPT Action or agent frameworkOpenAPI 3.1Pro API key

The remote MCP endpoint is always:

text
https://chartinspect.com/api/mcp

#OpenAI Responses API

ChartInspect is a remote MCP server, so the Responses API can discover and call its tools without converting them into function definitions. The value in authorization is your raw ChartInspect key, without a Bearer prefix. OpenAI sends it to the MCP server as a bearer credential.

ts
import OpenAI from 'openai';

const client = new OpenAI();
const response = await client.responses.create({
  model: 'gpt-5.6',
  input: 'Compare Bitcoin MVRV and NUPL over the last 365 days.',
  tools: [
    {
      type: 'mcp',
      server_label: 'chartinspect',
      server_description: 'Read-only crypto, on-chain, market, and economic data.',
      server_url: 'https://chartinspect.com/api/mcp',
      authorization: process.env.CHARTINSPECT_API_KEY,
      require_approval: 'never',
    },
  ],
});

console.log(response.output_text);

Set CHARTINSPECT_API_KEY=ci_live_your_pro_key in the server environment. OpenAI does not store the MCP authorization value, so include it on every Responses request. require_approval: 'never' is appropriate here because ChartInspect exposes only read-only discovery and data-fetch tools.

#ChatGPT custom app

ChatGPT can connect directly to the same remote MCP endpoint through a custom app:

  1. Enable developer mode in Settings → Apps → Advanced Settings.
  2. Create a custom app and enter https://chartinspect.com/api/mcp.
  3. Choose OAuth, scan the tools, then sign in to ChartInspect when prompted.

ChartInspect advertises OAuth discovery, PKCE, dynamic client registration, and refresh-token support. The offline_access scope is supported so compatible ChatGPT clients can keep the connection active after the one-hour access token expires. Availability of custom MCP apps depends on the user's ChatGPT plan and workspace settings.

#Claude.ai and Claude Desktop

Add a custom connector with this URL:

text
https://chartinspect.com/api/mcp

Claude sends you to ChartInspect to sign in and authorize read-only access. No API key is copied into Claude. The connection stops working if the ChartInspect Pro subscription lapses.

#Claude Messages API

Anthropic's current MCP connector uses an mcp_servers entry plus an mcp_toolset. Pass the raw ChartInspect key as authorization_token.

py
import anthropic

client = anthropic.Anthropic()
response = client.beta.messages.create(
    model='claude-opus-5',
    max_tokens=1200,
    messages=[
        {'role': 'user', 'content': 'Compare Bitcoin MVRV and NUPL over the last 365 days.'}
    ],
    mcp_servers=[
        {
            'type': 'url',
            'url': 'https://chartinspect.com/api/mcp',
            'name': 'chartinspect',
            'authorization_token': 'ci_live_your_pro_key',
        }
    ],
    tools=[{'type': 'mcp_toolset', 'mcp_server_name': 'chartinspect'}],
    betas=['mcp-client-2025-11-20'],
)

print(response)

#Claude Code, Cursor, and MCP clients

Claude Code can add the remote server in one command:

bash
claude mcp add --transport http chartinspect https://chartinspect.com/api/mcp --header "Authorization: Bearer ci_live_your_pro_key"

For clients that use an MCP JSON configuration:

json
{
  "mcpServers": {
    "chartinspect": {
      "type": "http",
      "url": "https://chartinspect.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ci_live_your_pro_key"
      }
    }
  }
}

The server uses Streamable HTTP. It supports MCP protocol versions 2025-06-18, 2025-03-26, and 2024-11-05.

#llms.txt

Fetch a compact, LLM-friendly map of the API and place it in an assistant's context:

bash
curl -H "X-API-Key: ci_live_your_pro_key" https://chartinspect.com/llms.txt

#OpenAPI and Custom GPT Actions

Fetch the OpenAPI 3.1 document:

bash
curl -H "X-API-Key: ci_live_your_pro_key" https://chartinspect.com/api/v1/openapi.json > chartinspect-openapi.json

For a Custom GPT Action, paste the fetched JSON into the Action schema editor and configure API-key authentication with the custom header X-API-Key. The schema URL itself is Pro-gated, so ChatGPT's unauthenticated URL importer cannot retrieve it directly. Use https://chartinspect.com/privacy-policy when the GPT editor asks for a privacy policy.

For new OpenAI API integrations, prefer the remote MCP setup above. It exposes four focused tools instead of turning every REST endpoint into a separate action.

#Security and limits

  • Read-only. The MCP server exposes discovery and data-fetch tools only.
  • Plan-bound. Access is checked live and stops when the Pro subscription lapses.
  • Quota-aware. Each get_metric call is a normal API request. Discovery calls do not consume the data quota.
  • Keep keys server-side. Never put an API key in browser code, an embed URL, or a public repository.
  • OAuth is keyless for the client. ChartInspect sees the signed-in ChartInspect account, not the user's OpenAI or Claude password.
Was this page helpful?