EarningsCalls MCP Server
Enables AI agents to access and search 177,000+ earnings call transcripts from 17,000+ companies, allowing natural language queries about financial commentary.
README
EarningsCalls MCP Server
Website · MCP setup guide · API docs · License: MIT
Model Context Protocol server that gives AI agents direct access to the EarningsCalls.dev dataset — 177,000+ earnings call transcripts from 17,000+ companies across 70 countries and 170+ exchanges (2020 to present), with 11M+ speaker segments cleanly tagged by role (Executive / Analyst / Operator / Attendee / Shareholder), full-text search, and more.
Ask Claude (or any MCP-capable client) about any earnings call in natural language:
"What did NVIDIA's CFO say about data center growth in their last call?" "Compare Tesla's gross-margin commentary across the last four quarters." "Find S&P 500 companies that mentioned 'tariffs' in their Q2 2026 calls."
Built as a thin MCP wrapper around the public REST API at https://earningscalls.dev/api/v1. New calls land within minutes of the call ending.
Requirements
- A paid EarningsCalls subscription (Pro, Ultra, or Enterprise). The free test key does not work here — grab a plan at earningscalls.dev/#pricing (from $24.99/month).
- For the local method only: Node.js 18+.
Installation
Two remote options (zero install, recommended) and one local option.
Option A — Connector URL (recommended)
The easiest path: a personal connector URL with the token embedded, so there's no API-key header to manage.
- Go to earningscalls.dev/dashboard → Connectors → Generate.
- Copy the URL — it looks like
https://earningscalls.dev/u/<your-token>/mcp.
Claude Desktop / claude.ai (web): Settings → Connectors → Add custom connector → paste the URL → Add. (On the web, leave the OAuth fields empty and confirm the consent popup.)
Claude Code (CLI):
claude mcp add --transport http earningscalls "https://earningscalls.dev/u/<your-token>/mcp"
# add --scope user to make it available in every project
Cursor — .cursor/mcp.json:
{
"mcpServers": {
"earningscalls": {
"url": "https://earningscalls.dev/u/<your-token>/mcp"
}
}
}
Each connector URL is independently revocable — generate one per device or workspace.
Option B — API-key header
Point your client at the hosted endpoint and pass your API key (ect_...) via the X-API-Key header.
Claude Code (CLI):
claude mcp add earningscalls --transport http https://mcp.earningscalls.dev/mcp \
--header "X-API-Key: ect_your_key_here"
Claude Desktop / Cursor — config:
{
"mcpServers": {
"earningscalls": {
"url": "https://mcp.earningscalls.dev/mcp",
"headers": { "X-API-Key": "ect_your_key_here" }
}
}
}
Option C — Local (via npx)
Runs the server as a local process over stdio.
claude mcp add earningscalls \
--env EARNINGSCALLS_API_KEY=ect_your_key_here \
-- npx -y @earningscalls/mcp-server
Or in a client config:
{
"mcpServers": {
"earningscalls": {
"command": "npx",
"args": ["-y", "@earningscalls/mcp-server"],
"env": { "EARNINGSCALLS_API_KEY": "ect_your_key_here" }
}
}
}
Coverage
| Earnings call transcripts | 177,000+ |
| Companies | 17,000+ |
| Countries / exchanges | 70 / 170+ |
| History | 2020 – present (5+ years) |
| Speaker segments | 11M+ (Executive / Analyst / Operator / Attendee / Shareholder) |
| Sectors | All 11 GICS sectors |
| Freshness | New calls within minutes of the call ending |
Available Tools
| Tool | Description |
|---|---|
get_dataset_stats |
Total counts and date range — call first to understand coverage. |
list_latest_calls |
Most recent calls, optionally filtered by sector. |
list_recent_transcripts |
Recently added transcripts. |
list_calls_by_ticker |
All calls for a ticker (e.g. PLTR, AAPL), newest first. |
get_latest_call_for_ticker |
Single most recent call for a ticker. |
get_company_by_name |
Look up a company by full/partial name. |
list_companies |
Browse / search companies. |
list_upcoming_earnings |
Scheduled calls in the next N days. |
get_earnings_call |
Metadata for a single call. |
get_transcript |
Full text of a call — format full, summary, or components. |
get_speaker_segments |
Speaker segments with role filter (Executive / Analyst / Operator …). |
search_transcripts |
Full-text search across all transcripts and speaker segments. |
search_within_ticker |
Full-text search scoped to one ticker. |
list_sectors / list_industries |
Discover filter dimensions. |
Environment Variables
Local mode (stdio)
| Variable | Required | Default |
|---|---|---|
EARNINGSCALLS_API_KEY |
yes | — |
EARNINGSCALLS_BASE_URL |
no | https://earningscalls.dev |
Remote mode (HTTP)
| Variable | Required | Default |
|---|---|---|
PORT |
no | 3000 |
EARNINGSCALLS_BASE_URL |
no | https://earningscalls.dev |
In remote mode the API key is provided per session by the connecting client — either embedded in the connector URL (Option A) or via the X-API-Key header (Option B).
Development
npm install
npm run build
npm start # local stdio mode
npm run serve # remote HTTP mode
Source layout:
src/
├── index.ts # stdio entry + preflight (local mode)
├── http.ts # Streamable HTTP entry (remote mode)
├── client.ts # HTTP client for earningscalls.dev API
├── config.ts # env var loading
├── messages.ts # user-facing banners + error hints
└── tools.ts # MCP tool definitions
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。