Talordata Mcp Server について
TalorData MCP Server は TalorData SERP 検索を MCP 対応のエージェントやクライアントに接続します。ライブ検索に加え、エンジンスキーマ、リクエスト履歴、利用統計を参照できます。
主な機能
- 01複数エンジンのライブ SERP 検索
- 02エンジン索引とスキーマリソース
- 03検索リクエスト履歴の照会
- 04利用統計の照会
- 05リクエストごとのユーザートークン転送
活用シーン
- 01MCP クライアントからライブ検索を実行
- 02スキーマを確認して検索パラメーターを構成
- 03過去の検索リクエストを確認
- 04日付やエンジン別の利用統計を確認
MCP 対応
TalorData SERP の検索、履歴、統計を MCP ツールとして公開し、エンジン索引とスキーマをリソースとして提供します。リモート HTTP 接続に対応し、各ユーザーのトークンを上流へ転送します。
設定
リモートサーバーはインストール不要です。クライアントを選んで設定をコピーするか、ワンクリックで導入できます。
https://your-domain.com:8800/mcpclaude mcp add --transport http talordata https://your-domain.com:8800/mcp{"mcpServers":{"talordata":{"type":"http","url":"https://your-domain.com:8800/mcp"}}}Cursor で開く{"servers":{"talordata":{"type":"http","url":"https://your-domain.com:8800/mcp"}}}VS Code で開く{"mcpServers":{"talordata":{"type":"http","url":"https://your-domain.com:8800/mcp"}}}{"mcpServers":{"talordata":{"serverUrl":"https://your-domain.com:8800/mcp"}}}コマンドはターミナルで実行し、JSON は各クライアントの MCP 設定に貼り付けて再起動します。認可が必要なサーバーは初回接続時にログインを求めます。
https://your-domain.com:8800/mcp認証が必要ツール
README(原文)English
プロジェクトの README よりこの README の日本語訳はまだありません。以下は英語の原文です。日本語の紹介はページ上部をご覧ください。
TalorData MCP Server
Connect TalorData SERP to MCP-compatible AI agents and clients for real-time search, usage history, and statistics.
TalorData SERP MCP Server exposes TalorData search capabilities through the Model Context Protocol (MCP), making it easy to integrate live search tools into agents, workflows, and MCP clients. In addition to real-time SERP requests, it also provides access to search history, usage statistics, and engine schema resources for richer integrations.
Quick Start
Get your API key
Sign up at TalorData and get your API key from the dashboard.
Run locally
git clone https://github.com/Talordata/talordata-mcp
cd talordata-mcp
go mod tidy
go run .
By default, configuration is loaded from configs/config.yaml.
The sample config in this repository is:
root_dir: .
listen_addr: ":8800"
upstream_endpoint: "https://serpapi.talordata.net/serp/v1/request"
history_endpoint: "https://api.talordata.com/accounts/v1/serp/mcp/history"
statistics_endpoint: "https://api.talordata.com/pay_package_view/v1/serp/mcp/statistics"
timeout_ms: 150000
shutdown_timeout_ms: 10000
log_prefix: "[talordata-mcp]"
After startup, the server exposes:
-
GET / -
GET /healthz -
POST | GET | DELETE /mcp -
POST | GET | DELETE /{user-token}/mcp
Compile only
go build ./...
Example MCP client config
Recommended remote HTTP MCP setup:
{
"mcpServers": {
"talordata": {
"url": "https://your-domain.com:8800/mcp",
"headers": {
"Authorization": "Bearer YOUR_USER_TOKEN"
}
}
}
}
For clients that cannot send custom headers:
{
"mcpServers": {
"talordata": {
"url": "https://your-domain.com:8800/YOUR_USER_TOKEN/mcp"
}
}
}
Features
-
Dynamically loads all engine schemas from local
engines/*.json -
Exposes engine index and raw engine schemas through MCP resources
-
Proxies Talor SERP search requests through the
searchtool -
Proxies usage history through the
historytool -
Proxies usage statistics through the
statisticstool -
Supports response format control via upstream
json -
Supports schema-aware parameter serialization rules
-
Supports per-request user token forwarding instead of storing a fixed upstream token
Supported serialization behavior
-
date_range -
tags -
cr -
switch -
time_range -
cascader -
number -
google_flightsairport code normalization
Response format mapping
-
json=1→ structured JSON only -
json=2→ JSON + HTML -
json=3→ HTML only
Runtime Model
-
Designed for cloud deployment with
streamable-http -
The server does not persist a shared upstream token
-
Each user provides their own Talor SERP token per MCP request
-
The server authenticates the incoming request and forwards the user token upstream
Authentication
User token
Supported token delivery methods:
-
Recommended:
Authorization: Bearer <user-token> -
Compatible:
X-Talor-Serp-Token: <user-token> -
Compatible:
/{user-token}/mcp
Notes
-
For
/mcp, sending the token in headers is recommended -
/{user-token}/mcpis useful for clients that cannot customize headers -
Query string token passing is intentionally not supported
-
The server does not infer
agent-platformfrom inbound requests
Configuration
Service configuration is loaded from configs/config.yaml.
Fields
| Field | Description |
|---|---|
root_dir | Project root directory; supports relative paths and must contain engines/index.json |
listen_addr | Service listen address |
upstream_endpoint | Talor SERP search endpoint |
history_endpoint | Talor SERP history endpoint |
statistics_endpoint | Talor SERP statistics endpoint |
timeout_ms | Upstream timeout in milliseconds |
shutdown_timeout_ms | Graceful shutdown timeout in milliseconds |
log_prefix | Log prefix used by the service |
Tools
Business tools are implemented under the tools directory:
-
tools/search.go -
tools/history.go -
tools/statistics.go
**list_engines**
Returns:
-
Default engine
-
Category list
-
Engine list
-
Schema resource URI for each engine
**search**
Executes a Talor SERP search request.
Parameters
| Parameter | Required | Description |
|---|---|---|
engine | No | Engine key such as google_web, google_images, bing_images |
q | No | Search query |
json | No | Response format: 1, 2, or 3 |
params | No | Engine-specific parameters |
response_mode | No | complete or compact |
Recommended flow
-
Read
talor://engines -
Read
talor://engines/<engine> -
Build
paramsaccording to the schema -
Call
search
**history**
Queries Talor SERP usage history.
Parameters
| Parameter | Required | Description |
|---|---|---|
page | No | Page number, default 1 |
page_size | No | Page size, commonly 20, 50, 100 |
search_query | No | Search query filter |
search_engine | No | Search engine filter |
status | No | all, success, or error |
start_time | No | Start time in Unix seconds |
end_time | No | End time in Unix seconds |
timezone | No | Forwarded as X-Time-Zone |
**statistics**
Queries Talor SERP usage statistics.
Parameters
| Parameter | Required | Description |
|---|---|---|
start_date | Yes | Start date in YYYY-MM-DD |
end_date | Yes | End date in YYYY-MM-DD |
engines | No | Comma-separated string or string array |
timezone | No | Timezone offset such as +08:00 |
Resources
| Resource | Description |
|---|---|
talor://engines | Engine index |
talor://engines/<engine> | Raw schema loaded from engines/<engine>.json |
Parameter Serialization Rules
When building upstream form parameters:
-
enginemust be set to the engine key -
jsonmust be included when needed by the upstream endpoint -
date_rangefields are expanded to{field}_startand{field}_end -
tagsvalues are joined with commas -
switchvalues are serialized as"true"/"false" -
cascaderuses the last selected value -
numbervalues are serialized as strings -
time_rangevalues are formatted asHHmm,HHm
Project Structure
| Path | Responsibility |
|---|---|
main.go | Service startup, MCP registration, HTTP routing, graceful shutdown |
tools/search.go | search tool definition and handler |
tools/history.go | history tool definition and handler |
tools/statistics.go | statistics tool definition and handler |
internal/auth/auth.go | Token extraction, auth middleware, MCP context injection |
internal/engines/registry.go | Engine index and schema loading |
internal/serp/client.go | Upstream HTTP requests |
internal/serp/serialize.go | Parameter serialization logic |
Development
Local checks
go build ./...
Health endpoints
-
GET /returns service metadata -
GET /healthzreturns health status
Learn more
Explore MCP and how to use it across different platforms:
コメント
AI アプリケーションが外部のツールやデータソースに統一的な方法で接続するためのオープンなプロトコルです。クライアントはモデルとの対話を担い、MCP サーバーはツール・リソース・プロンプトテンプレートを提供します。標準メッセージでやり取りするため、1 つのサーバーを複数のクライアントで再利用できます。