tokportal-mcp
MCP server for TokPortal's managed social infrastructure API, enabling creation, warming, and operation of real TikTok, Instagram, and YouTube accounts with content publishing.
README
tokportal
TokPortal is the managed social infrastructure API: real TikTok, Instagram and YouTube accounts created, warmed and operated by human account managers in 16+ countries — exposed as a REST API and an MCP server. No OAuth per account, no 25-posts/day cap, no app review.
Docs https://developers.tokportal.com · API base https://app.tokportal.com/api/ext · OpenAPI https://developers.tokportal.com/openapi.json · MCP remote https://app.tokportal.com/api/ext/mcp · Get an API key https://app.tokportal.com/developer/api-keys · llms.txt https://developers.tokportal.com/llms.txt
tokportal is the official Python SDK for the TokPortal API (Python 3.9+, typed, standard library only). Every public operation is available as a resource method or through the generated request_operation map.
Install
pip install tokportal
30-second quickstart
import os
from tokportal import TokPortal
client = TokPortal(api_key=os.environ["TOKPORTAL_API_KEY"])
# 1. Create a bundle: a fresh managed TikTok account in the USA + 1 video slot.
# Credits are debited now; the account manager is assigned at publish time.
bundle = client.bundles.create({
"bundle_type": "account_and_videos",
"platform": "tiktok",
"country": "USA",
"title": "US launch",
"videos_quantity": 1,
})
bundle_id = bundle["data"]["id"]
# 2. Upload the video straight from disk -> public_url
upload = client.uploads.video_direct("./launch.mp4", bundle_id, content_type="video/mp4")
# 3. Configure the account profile and video slot 1, then publish
client.bundles.configure_account(bundle_id, {
"username": "mybrand.us",
"visible_name": "My Brand",
"biography": "Official account",
})
client.bundles.configure_video(bundle_id, 1, {
"video_type": "video",
"video_url": upload["data"]["public_url"],
"description": "Day 1 - launching in the US #launch",
"target_publish_date": "2026-09-01",
})
client.bundles.publish(bundle_id)
# 4. Later (webhook `account.in_review` / `account.finalized`, or polling):
# saved_account_id is the real delivered account -> read it back
current = client.bundles.get(bundle_id)["data"]
if current.get("saved_account_id"):
account = client.accounts.get(current["saved_account_id"])["data"]
print(account["username"], account["profile_url"])
Method names follow the generated resource map (
bundles,uploads,accounts,analytics,webhooks). If a helper does not exist for an operation, useclient.request_operation("<operationId>", path=..., query=..., body=...).
Full example
import os
from tokportal import TokPortal, TokPortalApiError
client = TokPortal(api_key=os.environ["TOKPORTAL_API_KEY"])
me = client.me()
bundle = client.bundles.create({
"bundle_type": "account_and_videos",
"country": "USA",
"videos_quantity": 5,
})
csv = client.analytics.export_videos(account=["saved-account-id"])
image = client.uploads.image_from_url({
"url": "https://cdn.example.com/photo.jpg",
"bundle_id": bundle["data"]["id"],
})
print(me["data"]["email"], bundle["data"], csv, image["data"])
Direct multipart uploads use the same structured errors and idempotency support:
uploaded = client.uploads.video_direct(
"./video.mp4",
bundle["data"]["id"],
content_type="video/mp4",
idempotency_key="video-upload-123",
)
Manage TokPortal Coverage from the latest atomic quote. A zero-credit quote is valid and still requires an explicit reactivation call:
coverage = client.accounts.coverage("saved-account-id")
quote = coverage["data"]["reactivation_quote"]
if quote:
client.accounts.reactivate_coverage(
"saved-account-id",
{
"expected_credits": quote["credits"],
"expected_current_period_end": quote["current_period_end"],
"expected_lock_version": quote["lock_version"],
},
idempotency_key="coverage-reactivate-saved-account-id-v4",
)
client.accounts.pause_coverage(
"saved-account-id",
idempotency_key="coverage-pause-saved-account-id-v4",
)
Credential reveal and verification-code access use the same irreversible
two-step policy flow. First call without acceptance to receive HTTP 428 and
error.details.policy_version; then show those terms to the account owner and
retry with that exact version. The accepted request may debit credits and
permanently detach the account. These secret-bearing responses are never stored
for replay, so these helpers intentionally do not accept idempotency_key.
After an uncertain transport result, reconcile the safe account state before
deciding whether to call the endpoint again without a key:
If an accepted call returns HTTP 409 with
CREDENTIAL_REVEAL_QUOTE_CHANGED, no charge or reveal occurred. Read the
current policy and expected_credit_cost from error.details, show the new
terms to the owner, obtain fresh consent, and retry with the new version. Never
retry a 409 automatically.
try:
client.accounts.reveal_credentials("saved-account-id")
except TokPortalApiError as error:
if error.status_code != 428:
raise
policy_version = str(error.details["policy_version"])
credentials = client.accounts.reveal_credentials(
"saved-account-id",
acceptance={
"acknowledge_support_forfeit": True,
"policy_version": policy_version,
},
)
The same no-replay rule applies to webhooks.create, uploads.image,
uploads.video, and analytics.create_report because they return a signing
secret, signed upload capability, or report access token. These helpers do not
accept idempotency_key, and request_operation rejects one locally for all
six sensitive operation IDs.
Discover and operate webhooks without dropping to raw HTTP:
catalog = client.webhooks.events()
endpoints = client.webhooks.list(event="bundle.published")
retry = client.webhooks.retry_delivery(endpoints["data"][0]["id"], "delivery-id")
Every OpenAPI operation is also reachable through the generated operation map:
same_retry = client.request_operation(
"retryWebhookDelivery",
path={"id": endpoints["data"][0]["id"], "delivery_id": "delivery-id"},
)
csv_again = client.request_operation(
"exportAnalyticsVideos",
query={"account": ["saved-account-id"]},
)
The SDK sends X-TokPortal-Client: tokportal-python/0.1.0 on API requests for observability and support diagnostics.
Verify signed webhook deliveries with the exact raw request body:
from tokportal import verify_webhook_signature
valid = verify_webhook_signature(
raw_body,
request.headers["TokPortal-Signature"],
os.environ["TOKPORTAL_WEBHOOK_SECRET"],
)
from tokportal import TokPortalApiError
try:
client.bundles.create({
"bundle_type": "account_and_videos",
"country": "USA",
"videos_quantity": 5,
})
except TokPortalApiError as error:
print(error.status_code, error.code, error.details, error.request_id)
if error.retryable:
wait_seconds = error.retry_after_seconds or 1
# Retry with backoff.
pass
print(error.rate_limit)
API keys use the format sk_ followed by 64 lowercase hex characters. TokPortal stores only a SHA-256 hash of the key and shows the raw key once at creation.
Source of truth
This package is generated from the TokPortal public OpenAPI schema
(https://developers.tokportal.com/openapi.json) in the private TokPortal
monorepo. Generated files (tokportal/_generated.py) are overwritten on every release — do not edit
them by hand. See CONTRIBUTING.md for what we accept as PRs
and SECURITY.md for vulnerability reporting.
Links
- Documentation: https://developers.tokportal.com
- SDKs & CLI guide: https://developers.tokportal.com/sdks-cli
- MCP server: https://developers.tokportal.com/mcp ·
tokportal-mcp - API reference (OpenAPI): https://developers.tokportal.com/openapi.json
- Other packages:
@tokportal/node·@tokportal/cli·tokportal(PyPI) ·github.com/tokportal/tokportal-go
MIT © TokPortal
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。