> ## Documentation Index
> Fetch the complete documentation index at: https://api-docs.ollang.com/llms.txt
> Use this file to discover all available pages before exploring further.
# How It All Connects
> How Ollang APIs, the MCP server, the SDK, Skills, and platform workflows fit together — and when to use each.
# How It All Connects
Ollang exposes the same underlying platform through four integration surfaces. They are complementary, not exclusive — most teams use two or three of them together.
```
┌──────────────────────┐
│ Olabs Dashboard │ ← humans: create orders, configure workflows, review
│ (lab.ollang.com) │
└──────────────────────┘
│
▼
┌──────────────────────────────────────────────────────────────┐
│ Ollang Integration API (api-integration.ollang.com)│
│ REST endpoints for uploads, orders, projects, QC, revisions, │
│ human review, custom instructions, folders, callbacks. │
└──────────────────────────────────────────────────────────────┘
▲ ▲ ▲ ▲
│ │ │ │
│ │ │ │
┌──────┴─────┐ ┌───────┴──────┐ ┌───────┴──────┐ ┌──────┴──────┐
│ Your code │ │ Ollang MCP │ │ Ollang SDK │ │ Ollang │
│ (curl, │ │ (Cloudflare, │ │ (npm, │ │ Skills │
│ HTTP libs)│ │ OAuth 2) │ │ TypeScript) │ │ (file-based)│
└────────────┘ └────────────────┘ └───────────────┘ └─────────────┘
```
## The Four Integration Surfaces
Programmatic access. Works in any language with an HTTP client. API-key authenticated.
Hosted Model Context Protocol endpoint. Plug into Claude Desktop, Cursor, Claude Code, Devin, Replit, Windsurf, and other MCP clients. OAuth 2.0 + PKCE.
TypeScript/Node.js library (`@ollang-dev/sdk`) for asset scanning, i18n workflows, CMS capture, and a typed REST client.
File-based [Agent Skills](https://skills.sh) bundle that teaches your AI coding agent how to call Ollang. No proxy server, no MCP needed.
## When to Use Which
| Scenario | Best fit |
| ------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------- |
| Build a translation pipeline from a backend or CI job | **REST API** (any language) or **SDK** (if Node.js) |
| Let translators / PMs trigger work from their AI assistant (Claude, Cursor, Devin) without writing code | **MCP** |
| Localize an internal Next.js / React / Vue app's i18n files | **SDK** Asset Management |
| Localize CMS content (Strapi, custom) captured live in the browser | **SDK** Browser SDK |
| Add translation to an existing AI coding workflow without running a server | **Skills** |
| Trigger orders from no-code/low-code tools (Airtable, Notion) | **MCP** with the matching [integration](/mcp/integrations/airtable) |
| Run Ollang as part of a workflow editor / agent runtime | **MCP** (preferred) or **REST API** if you can't run MCP |
## Authentication at a Glance
| Surface | Auth |
| ---------- | --------------------------------------------------------------------------------------------------- |
| REST API | API key in `X-Api-Key` header. Generate one in the [Olabs Dashboard](https://lab.ollang.com). |
| MCP Server | OAuth 2.0 + PKCE. Sign in with email OTP, Google, or GitHub the first time you connect. |
| SDK | API key (via `OLLANG_API_KEY` env var or constructor argument). |
| Skills | API key. The agent reads `OLLANG_API_KEY` from your environment or asks you to provide one in chat. |
All MCP traffic still flows through the same Integration API behind the
scenes — MCP is a protocol layer that translates conversational tool calls
into Ollang REST calls.
## A Concrete End-to-End Example
A typical localization flow — phrased four ways:
```bash theme={null}
# 1. Upload
curl -X POST "https://api-integration.ollang.com/integration/upload/direct" \
-H "X-Api-Key: " \
-F "file=@sample.mp4" \
-F "name=sample" \
-F "sourceLanguage=en"
# → { "projectId": "" }
# 2. Create subtitle order for French + Spanish
curl -X POST "https://api-integration.ollang.com/integration/orders/create" \
-H "X-Api-Key: " \
-H "Content-Type: application/json" \
-d '{
"orderType": "subtitle",
"level": 0,
"projectId": "",
"targetLanguageConfigs": [{"language":"fr"},{"language":"es"}],
"autoQc": true
}'
# → [{ "orderId": "" }, { "orderId": "" }]
```
```typescript theme={null}
import Ollang from "@ollang-dev/sdk";
const ollang = new Ollang({
apiKey: process.env.OLLANG_API_KEY,
});
const { projectId } = await ollang.uploads.uploadFile({
file: "./sample.mp4",
name: "sample",
sourceLanguage: "en",
});
const orders = await ollang.orders.create({
projectId,
orderType: "subtitle",
level: 0,
targetLanguageConfigs: [{ language: "fr" }, { language: "es" }],
autoQc: true,
});
```
```
Upload https://example.com/sample.mp4 to Ollang as English, then create
a subtitle order for French and Spanish with auto QC enabled. Report
the order IDs.
```
The AI assistant calls `uploadFile` → `createOrder` on your behalf and prints the resulting `orderId`s. Large files such as videos upload from a URL; smaller local files (e.g., PDF or DOCX documents) can be passed directly as base64-encoded content.
In all four cases:
1. The same upload endpoint runs (`POST /integration/upload/direct`).
2. The same order endpoint runs (`POST /integration/orders/create`).
3. The same project / order records are created in your Ollang account.
4. The same callback fires (if you set one).
## Workflows Apply Equally Everywhere
The **workflow** layer (which AI providers run, which review gates apply, which custom instructions are used) is configured in the [Olabs Dashboard](https://lab.ollang.com) at the Global or Folder level. It applies the same way no matter which surface created the order. See [Workflows and Provider Architecture](/workflows-provider-architecture).
This means:
* A developer creating an order via the REST API gets the same provider routing and review-gate behavior as a PM creating one in the dashboard, or an AI agent creating one via MCP.
* Changing a folder workflow affects all future orders into that folder, regardless of where they originate.
## Picking a Path — Quick Recommendations
* **Just exploring?** Run through [Getting Started](/getting-started) with `curl`.
* **Shipping production code?** Use the [REST API](/apis/ollang-api-reference/direct-file-upload) or, on Node.js, the [SDK](/sdk/overview).
* **AI-driven team workflows?** Connect [MCP](/mcp/overview) once and any compatible client can use Ollang.
* **AI-driven developer workflows in your editor?** [Skills](/skills/overview) — one command, no server.