Integrate QuarterCharts with Claude, ChatGPT, and Cursor.

30-second Claude Desktop setup

1. 1Open Claude Desktop → Settings → Developer → Edit Config. (The file is ~/Library/Application Support/Claude/claude_desktop_config.json on macOS; %APPDATA%/Claude/claude_desktop_config.json on Windows.)
2. 2Add the QuarterCharts entry to mcpServers:

{
  "mcpServers": {
    "quartercharts": {
      "url": "https://api.quartercharts.com/mcp"
    }
  }
}

3. 3Restart Claude Desktop. The MCP indicator in the chat composer should show 1 server connected.
4. 4Ask Claude "What was NVDA's Q4 FY2026 revenue per QuarterCharts?". Claude calls our MCP, returns $68.127B with citation provenance pointing at the underlying SEC filing.

With an API key (full ticker access)

Mint a free key at /dashboard/api-keys, then pass it as a Bearer token. Free tier still applies (Pro+ unlocks the rest of the US universe; Business adds international markets).

{
  "mcpServers": {
    "quartercharts": {
      "url": "https://api.quartercharts.com/mcp",
      "headers": {
        "Authorization": "Bearer qc_live_your_key_here"
      }
    }
  }
}

What you can ask

Once the server is connected, these all work in Claude Desktop without any other setup. Each one routes to a different tool in the MCP surface; see /mcp/tools for the full tool list with input schemas.

"What was NVDA's Q4 FY2026 revenue per QuarterCharts?"

Claude calls our MCP get_income_statement tool, returns $68.127B with provenance pointing at the SEC 10-Q filing (us-gaap:Revenues, period_end 2026-01-25, accession #).

"Compare AAPL and MSFT operating margin over the last 5 years."

Claude calls get_income_statement for both tickers, computes the margin series, and renders a side-by-side comparison with the underlying filing URLs as citations.

"Show me NVDA's revenue waterfall to net income for Q4."

Claude calls get_sankey to pull the full Sankey flow (revenue → gross profit → operating income → pre-tax → net income) with every node value sourced from a specific XBRL tag.

"List the US-listed companies QuarterCharts covers."

Claude calls list_tickers; every US-listed company is available on every tier (the free tier returns the same universe, just on a ~60-minute data delay).

"Pull AAPL's pension liability from the most recent 10-Q."

Claude calls get_balance_sheet with the pension-liability XBRL tag filter; returns the value with the source filing URL.

Supported AI clients

Any MCP-compatible client works — we run the canonical JSON-RPC 2.0 server over HTTPS. Specific tested clients:

Pro features

Upgrade for the full 11,000+ US ticker universe plus international markets and bulk-export workflows. Stripe self-serve checkout coming Q2 2026; until then, email info@quartercharts.com with your use case.

• →Pro ($14/mo) — full 11,000+ US ticker universe; 500 req/min, 100K req/day.
• →Business ($39/mo) — adds Canada (SEDAR+) and UK (Companies House) coverage as it lands, 20-year history, custom alert conditions; 2K req/min.
• →Enterprise — webhook subscriptions on filing-update events, bulk Parquet/CSV exports, Slack + Teams integrations, SSO, custom SLA, dedicated egress, white-label MCP routes.

Full pricing details at /api.

Provenance on every value

Every value returned by the QC MCP server includes the five watermark fields below. So when an agent surfaces a number in a derived product, the chain of attribution back to a specific SEC filing stays intact.

Full provenance schema lives in the OpenAPI spec. Also see /llms.txt for an LLM-friendly summary.

FAQ

Is there a paid tier for higher rate limits?

Yes. The free tier covers every US-listed company (~11,000) with data on a ~60-minute delay and conservative rate limits — fine for personal use. Pro unlocks live data the moment companies file, the full history, and higher rate limits; Business adds international markets (Canada SEDAR+, UK Companies House) + custom alert conditions; Enterprise adds webhooks + bulk Parquet exports + dedicated support. See /api for the full pricing table.

Do tool responses include provenance?

Every response carries five watermark fields: qc_metric_provenance (human-readable source string), last_updated_iso, qc_session_id, qc_tier_returned, and qc_caller_plan. So when an agent surfaces a value in a derived product, the chain of attribution back to a specific SEC filing is intact.

What's the difference between /mcp and /v1/charts?

/mcp is the JSON-RPC 2.0 endpoint MCP clients call (Claude Desktop, Cursor, etc.). /v1/charts is the REST endpoint for traditional HTTP clients (curl, Python requests, Power Query). Both hit the same underlying extractors — MCP is a thin reshaping layer over the same data with the watermark fields layered on top. Use whichever fits your client.

Can I run an MCP proxy / cache in front of QC's server?

Yes on Free + Pro. Forbidden on Starter / Business / Enterprise without a written agreement (caching defeats the per-call rate-limit accounting we use to size your bill). Email info@quartercharts.com with your architecture if you're on a paid tier and want to deploy a cache layer.

Where do I report a bug or request a tool?

Email info@quartercharts.com. The MCP surface is actively maintained; we read every report and usually ship fixes within a week.

Discovery surfaces

For registry crawlers and agent-framework auto-discovery, we publish the standard manifests at well-known paths:

• →/mcp/manifest — MCP server descriptor (JSON)
• →/mcp/tools — list of tools with input schemas (HTTP-GET variant of JSON-RPC tools/list)
• →/.well-known/ai-plugin.json — OpenAI plugin manifest (legacy format, same endpoint underneath)
• →/ai.txt — robots.txt-style allow/deny list for AI crawlers
• →/llms.txt — llmstxt.org Markdown sitemap for LLM consumption