集成
身份和欺诈的最佳方式。

# Integrate Didit into my application

You are integrating Didit — infrastructure for identity and fraud — into my application end-to-end. Didit ships KYC, KYB, AML screening, biometric verification, transaction monitoring, and wallet screening behind one API. 220+ countries, 14,000+ document types, 48+ languages. Pay-per-call from $0.33, 500 free verifications per core feature per month, no minimums, no contract.

## My application context

<my_stack>
Stack: [framework + language — e.g. Next.js 15 + TypeScript, Django, Rails, Go]
Surface: [web | iOS | Android | React Native | Flutter | backend-only]
Use case: [KYC at signup | KYB at business onboarding | age gate | re-auth | transaction monitoring | other]
Database: [Postgres | MySQL | Mongo | other]
</my_stack>

## Step 1 — Create the Didit account programmatically

Two API calls. No browser. No 2FA. Returns an `api_key` you use for every subsequent request.

```bash
# 1) Register — creates org + app + emails a 6-character code
curl -X POST https://apx.didit.me/auth/v2/programmatic/register/   -H "Content-Type: application/json"   -d '{"email": "dev@example.com", "password": "StrongP@ss1"}'

# 2) Verify email with the code from the inbox — returns api_key
curl -X POST https://apx.didit.me/auth/v2/programmatic/verify-email/   -H "Content-Type: application/json"   -d '{"email": "dev@example.com", "code": "A3K9F2"}'
# Response.application.api_key — save this
```

Persist the returned `api_key` as the env var `DIDIT_API_KEY`. Auth header for every Didit call:
`x-api-key: $DIDIT_API_KEY`

## Base URLs

| Surface | Base URL |
|---|---|
| Every `/v3/...` endpoint (sessions, workflows, lists, billing, webhook destinations, entities, transactions, standalone APIs) | `https://verification.didit.me` |
| Auth / programmatic registration (`/auth/v2/programmatic/...`) | `https://apx.didit.me` |
| Console (human UI) | `https://business.didit.me` |
| Hosted verification URL handed to end users | `https://verify.didit.me` |
| Docs (cross-check anything) | `https://docs.didit.me` |
| OpenAPI spec (machine-readable, server is `https://verification.didit.me`) | `https://docs.didit.me/openapi-25.json` |

## Step 2 — Choose the integration approach

Pick exactly one, based on my use case:

- **Approach A — Sessions API + SDK (recommended for any end-user verification flow).** My backend creates a Didit session → my frontend opens the Didit-hosted verification UI (via SDK / iframe / redirect) → user completes verification → my backend receives a webhook with the decision. Best for KYC at signup, KYB at business onboarding, age verification, re-auth. Didit-hosted flows are A/B-tested and convert higher than custom UIs.
- **Approach B — Standalone APIs (server-to-server only).** My backend calls individual modules directly (`/v3/id-verification/`, `/v3/aml/`, `/v3/face-match/`, etc.) for batch jobs or fully custom UI. Best for back-office verification pipelines.

If unsure, default to Approach A.

## Step 3 — Create or reuse a workflow

A workflow defines which modules run during a session. Create one via API (or `https://business.didit.me` → Workflows → Create).

```bash
# List existing workflows
curl https://verification.didit.me/v3/workflows/   -H "x-api-key: $DIDIT_API_KEY"

# Create a new KYC workflow — features are UPPERCASE enum values, each an object with optional `config`
curl -X POST https://verification.didit.me/v3/workflows/   -H "x-api-key: $DIDIT_API_KEY"   -H "Content-Type: application/json"   -d '{
    "workflow_label": "Standard KYC",
    "features": [
      { "feature": "OCR" },
      { "feature": "LIVENESS", "config": { "face_liveness_method": "PASSIVE" } },
      { "feature": "FACE_MATCH" },
      { "feature": "IP_ANALYSIS" }
    ]
  }'
# Response.uuid — save as DIDIT_WORKFLOW_ID
```

Feature enum values (UPPERCASE, source: `management-api/workflows/feature-configs.mdx`):
- **KYC features:** `OCR`, `NFC`, `LIVENESS`, `FACE_MATCH`, `AGE_ESTIMATION`, `PHONE_VERIFICATION`, `EMAIL_VERIFICATION`, `DATABASE_VALIDATION`, `AML`, `IP_ANALYSIS`, `PROOF_OF_ADDRESS`, `QUESTIONNAIRE`.
- **KYB features:** `KYB_REGISTRY`, `KYB_DOCUMENTS`, `KYB_KEY_PEOPLE` (each plus AML / DATABASE_VALIDATION as needed).

For a KYB workflow set `"workflow_type": "kyb"` (default is KYC). A workflow is locked to its type at creation; create separate workflows per kind.

## Step 4 — Install the SDK that matches my stack

| Stack | Package | Install |
|---|---|---|
| Web (React, Vue, Next.js, Nuxt, Svelte, vanilla JS) | `@didit-protocol/sdk-web` | `npm install @didit-protocol/sdk-web` |
| iOS (Swift / SwiftUI) | `didit-sdk-ios` | SPM: `https://github.com/didit-protocol/didit-sdk-ios` |
| Android (Kotlin / Jetpack Compose) | `me.didit:didit-sdk` | Maven |
| React Native (Expo or bare) | `@didit-protocol/sdk-react-native` | `npm install @didit-protocol/sdk-react-native` |
| Flutter | `didit_sdk` | `flutter pub add didit_sdk` |
| Backend-only / batch / custom UI | none — call REST directly | — |

Always prefer native iOS / Android SDKs over WebView on mobile (NFC + camera + biometrics work natively).

## Step 5 — Create a session and present it to the user

**Backend creates the session:**

```bash
curl -X POST https://verification.didit.me/v3/session/   -H "x-api-key: $DIDIT_API_KEY"   -H "Content-Type: application/json"   -d '{
    "workflow_id": "'"$DIDIT_WORKFLOW_ID"'",
    "vendor_data": "internal-user-id",
    "callback": "https://myapp.com/done"
  }'
```

Response (201 Created):
```json
{
  "session_id": "4c5c7f3a-...",
  "session_token": "eyJ...",
  "url": "https://verify.didit.me/session/...",
  "status": "Not Started",
  "session_kind": "user",
  "workflow_id": "...",
  "vendor_data": "internal-user-id"
}
```

`vendor_data` is **my internal user ID** — Didit links the session to a User entity (auto-created if new). For KYB workflows it links to a Business entity.

Optional create-session field: `callback_method` (`"initiator"` | `"completer"` | `"both"`, default `"initiator"`) — controls which side of a cross-device flow receives the `callback` redirect. Useful when the verification opens on a phone after a desktop start.

**Frontend presents the verification (pick one):**

| Pattern | Code |
|---|---|
| Web JS SDK (modal) | `import { DiditSdk } from "@didit-protocol/sdk-web"; DiditSdk.shared.startVerification({ url })` |
| Iframe (embedded) | `<iframe src={url} allow="camera; microphone; fullscreen; autoplay; encrypted-media" />` |
| Redirect (cross-device) | `window.location.href = url` |
| iOS / Android / RN / Flutter | `DiditSdk.startVerification(token: session_token)` |

## Step 6 — Set up the webhook to receive results

Build `POST /api/webhooks/didit` in my backend.

**Register the webhook destination once:**

```bash
curl -X POST https://verification.didit.me/v3/webhook/destinations/   -H "x-api-key: $DIDIT_API_KEY"   -H "Content-Type: application/json"   -d '{
    "url": "https://myapp.com/api/webhooks/didit",
    "label": "production",
    "subscribed_events": ["status.updated", "data.updated"]
  }'
```

Response includes `secret_shared_key` — save as `DIDIT_WEBHOOK_SECRET`.

**Cloudflare / restrictive firewall users:** allowlist `18.203.201.92` — Didit webhooks egress from this single IP.

**Three signature headers** ship on every webhook. Verify **one** — `X-Signature-V2` is recommended because it survives JSON middleware re-encoding:

| Header | What it signs | When to use |
|---|---|---|
| `X-Signature-V2` ★ recommended | `JSON.stringify(sortKeys(shortenFloats(parsed_body)))` with unescaped Unicode | Default — works through Express / Django / FastAPI / Next.js body parsers |
| `X-Signature` | The raw request bytes verbatim | Only if you can guarantee no middleware re-encodes the body |
| `X-Signature-Simple` | `"{timestamp}:{session_id}:{status}:{webhook_type}"` | Fallback when nothing else works — does NOT verify the `decision` payload, so only use as a hint |

**Endpoint requirements (in order):**

1. Parse JSON (V2 is parser-tolerant; you can use any framework's default body parser).
2. Read headers: `X-Signature-V2` (HMAC-SHA256 hex), `X-Timestamp` (Unix seconds).
3. Reject if `abs(now - X-Timestamp) > 300` (replay protection).
4. Apply `shortenFloats` (whole-number floats → integers), then `sortKeys` (recursive lexicographic), then `JSON.stringify` (unescaped Unicode — default).
5. Compute `expected = HMAC-SHA256(DIDIT_WEBHOOK_SECRET, canonical_json, "utf8")`.
6. Compare with `X-Signature-V2` using constant-time comparison (`hmac.compare_digest` in Python, `crypto.timingSafeEqual` in Node).
7. Dispatch on `webhook_type`. Return `200` immediately even if processing is async — Didit retries 5xx/404 twice (~1 min, ~4 min).

**Node.js / Next.js App Router example (canonical V2):**

```ts
import crypto from "node:crypto";

// Server-side normalisation: convert whole-number floats to integers.
function shortenFloats(v: unknown): unknown {
  if (Array.isArray(v)) return v.map(shortenFloats);
  if (v && typeof v === "object") {
    return Object.fromEntries(
      Object.entries(v as Record<string, unknown>).map(([k, x]) => [k, shortenFloats(x)])
    );
  }
  if (typeof v === "number" && !Number.isInteger(v) && v % 1 === 0) return Math.trunc(v);
  return v;
}

// Recursively sort object keys (arrays preserved in order).
function sortKeys(v: unknown): unknown {
  if (Array.isArray(v)) return v.map(sortKeys);
  if (v && typeof v === "object") {
    return Object.keys(v as object)
      .sort()
      .reduce<Record<string, unknown>>((acc, k) => {
        acc[k] = sortKeys((v as Record<string, unknown>)[k]);
        return acc;
      }, {});
  }
  return v;
}

export async function POST(req: Request) {
  const raw = await req.text();
  const sig = req.headers.get("x-signature-v2") ?? "";
  const ts = Number(req.headers.get("x-timestamp"));
  if (!ts || Math.abs(Date.now() / 1000 - ts) > 300)
    return new Response("stale", { status: 401 });

  const parsed = JSON.parse(raw);
  const canonical = JSON.stringify(sortKeys(shortenFloats(parsed)));
  const expected = crypto
    .createHmac("sha256", process.env.DIDIT_WEBHOOK_SECRET!)
    .update(canonical, "utf8")
    .digest("hex");
  if (
    sig.length !== expected.length ||
    !crypto.timingSafeEqual(Buffer.from(expected), Buffer.from(sig))
  )
    return new Response("bad sig", { status: 401 });

  // parsed.webhook_type → dispatch …
  return new Response("ok");
}
```

**Webhook event types:** `status.updated`, `data.updated`, `user.status.updated`, `user.data.updated`, `business.status.updated`, `business.data.updated`, `activity.created`, `transaction.created`, `transaction.status.updated`.

**Webhook body shape:**
```json
{
  "session_id": "uuid",
  "status": "Not Started | In Progress | Awaiting User | In Review | Approved | Declined | Resubmitted | Abandoned | Expired | Kyc Expired",
  "webhook_type": "status.updated",
  "timestamp": 1627680000,
  "workflow_id": "uuid",
  "vendor_data": "internal-user-id",
  "metadata": { "any": "json" },
  "decision": { /* present when Approved / Declined / In Review; absent on Expired / Abandoned / Not Started */ }
}
```

The `decision` object holds per-module **plural arrays** (V3 schema — each entry keyed by its `node_id` in the workflow):
- `id_verifications[]`: first_name, last_name, document_type, document_number, date_of_birth, nationality, expiration_date, issuing_state, address, parsed_address, mrz, front_image_quality_score, back_image_quality_score, warnings[]
- `nfc_verifications[]`: status, document_type, mrz, signature_status, chip_data
- `liveness_checks[]`: status, score (0–1), method ("passive" | "active"), reference_image
- `face_matches[]`: status, score (0–100), source_image, target_image, warnings[]
- `phone_verifications[]`: status, phone_number, carrier, is_disposable, is_virtual
- `email_verifications[]`: status, email, is_breached, is_disposable, is_undeliverable, breaches[]
- `poa_verifications[]`: status, document_type, issuer, poa_address, poa_parsed_address, issue_date, expiration_date
- `aml_screenings[]`: status, total_hits, hits[] (pep_matches, sanction_matches, warning_matches, adverse_media_matches), entity_type ("person" | "company")
- `ip_analyses[]`: status, ip_address, country, vpn, proxy, tor, hosting, risk_score
- `database_validations[]`: status, provider, match_type, fields_matched
- `questionnaire_responses`: nullable object keyed by question_id → answer

V2 → V3 migration note: V2 used singular keys (`aml`, `phone`, `email`, `poa`). V3 renamed them to plural arrays so a single workflow can run the same module multiple times on different nodes — index by `node_id`.

## Step 7 — Apply the decision to my database

```ts
switch (event.status) {
  case "Approved":       user.verified = true; user.verifiedAt = new Date(); storeDecision(event.decision); break;
  case "Declined":       user.verificationStatus = "declined"; logWarnings(event.decision); break;
  case "In Review":      user.verificationStatus = "pending_review"; break;
  case "In Progress":    user.verificationStatus = "in_progress"; break;
  case "Awaiting User":  user.verificationStatus = "awaiting_user"; break; // KYB only — waiting for a UBO / officer KYC sub-session
  case "Resubmitted":    user.verificationStatus = "resubmitted"; break; // reviewer asked the user to retry
  case "Abandoned":      scheduleReminderEmail(user); break;
  case "Expired":        break; // session URL aged out before the user finished
  case "Kyc Expired":    user.verified = false; createNewSession(user); break; // verified user's KYC has aged out per retention policy
  case "Not Started":    break;
}
```

Idempotency:
- Session webhooks: dedupe on `session_id + webhook_type + timestamp`.
- Transaction webhooks (`transaction.created`, `transaction.status.updated`): dedupe on the `event_id` field — multiple events can target the same transaction.

Didit retries 5xx/404 twice (~1 min, ~4 min after the prior failure). Return 200 even if processing is async.

## Module catalogue (use whichever ones match my use case)

| Module | Endpoint | Price | Free tier |
|---|---|---|---|
| Core KYC bundle (ID + Liveness + Face Match + IP) | session w/ workflow | **$0.33** | 500/mo (per feature) |
| ID Verification (standalone) | `POST /v3/id-verification/` | $0.20 | 500/mo |
| Passive Liveness (standalone) | `POST /v3/passive-liveness/` | $0.05 | 500/mo |
| Active Liveness | session-only | $0.15 | — |
| Face Match 1:1 | `POST /v3/face-match/` | $0.05 | 500/mo |
| Face Search 1:N | `POST /v3/face-search/` | $0.05 | — |
| Age Estimation | `POST /v3/age-estimation/` | $0.10 | — |
| AML Screening | `POST /v3/aml/` | $0.20 | — |
| Ongoing AML Monitoring | org-level continuous (per active user/yr) | $0.07 / user / year | — |
| Database Validation (gov registries) | `POST /v3/database-validation/` | variable | — |
| Proof of Address | `POST /v3/poa/` | $0.20 | — |
| Email Verification | `POST /v3/email/send/` + `POST /v3/email/check/` | $0.03 | — |
| Phone Verification (SMS / WhatsApp / voice / RCS / Telegram) | `POST /v3/phone/send/` + `POST /v3/phone/check/` | $0.04 + carrier | — |
| NFC Verification (native SDKs only) | session module | $0.15 | — |
| Biometric Authentication (re-auth) | session module | $0.10 | — |
| Device & IP Analysis | session module | $0.03 | — |
| Custom Questionnaire | session module | $0.10 | — |
| Business Verification (KYB) | `POST /v3/session/` w/ KYB workflow | $2.00 | — |
| Company AML Screening | KYB module | $0.20 | — |
| Person AML (per UBO / officer) | KYB module | $0.20 | — |
| Transaction Screening (rule engine) | `POST /v3/transactions/` | $0.02 / tx | — |
| AML Transaction Screening (counterparty sanctions + on-chain wallet risk) | inside `/v3/transactions/` | $0.15 / tx | — |
| Reusable KYC (share verified user across partners) | session feature | Free | — |
| White Label | workflow option | $0.20 | — |

## Operational APIs your code should know

| Action | Endpoint |
|---|---|
| Read full decision JSON | `GET /v3/session/{id}/decision/` |
| List sessions (filter by status, workflow, vendor_data, date) | `GET /v3/sessions` |
| Manually approve / decline / request review | `PATCH /v3/session/{id}/update-status/` |
| Patch metadata or extracted data | `PATCH /v3/session/{id}/update-data/` |
| Download compliance PDF | `GET /v3/session/{id}/generate-pdf` |
| Delete a session (GDPR / cleanup) | `DELETE /v3/session/{id}/delete/` |
| Share session with a partner (Reusable KYC) | `POST /v3/session/{id}/share/` |
| Submit a transaction for monitoring | `POST /v3/transactions/` |
| List / create lists (blocklist, allowlist, custom) | `GET/POST /v3/lists/` |
| Add entry to a list | `POST /v3/lists/{id}/entries/` |
| Upload a face image to a blocklist | `POST /v3/lists/{id}/entries/face-upload/` |
| Check credit balance | `GET /v3/billing/balance/` |
| Top up credits (returns Stripe checkout URL) | `POST /v3/billing/top-up/` |
| List / update / delete vendor users | `GET/PATCH /v3/users/`, `/v3/users/{vendor_data}/` |
| List / update / delete vendor businesses | `GET/PATCH /v3/businesses/` |

Every `/v3/...` endpoint above is served from `https://verification.didit.me` (the canonical OpenAPI server) and authenticates with `x-api-key`. Auth/registration is the only surface on `https://apx.didit.me`.

## Best-practice checklist before you mark this done

- [ ] `DIDIT_API_KEY`, `DIDIT_WORKFLOW_ID`, `DIDIT_WEBHOOK_SECRET` in env. Never committed.
- [ ] Backend route that creates a session for a given user, returns `url` (the hosted verification URL on `verify.didit.me`) and the `session_id`.
- [ ] Frontend that opens the session `url` via the right SDK / iframe / redirect for my stack.
- [ ] Webhook endpoint with: `X-Timestamp` freshness check (≤ 300s), canonical-V2 JSON re-serialisation (`shortenFloats` → `sortKeys` → unescaped `JSON.stringify`) HMAC-SHA256, constant-time compare against `X-Signature-V2`, dispatch on `webhook_type`, return 200.
- [ ] DB schema + update logic for `status in (Not Started, In Progress, Awaiting User, In Review, Approved, Declined, Resubmitted, Abandoned, Expired, Kyc Expired)`.
- [ ] Idempotent webhook handler (dedupe on `session_id + webhook_type`).
- [ ] No API key shipped to the browser. The session creation is server-side only.
- [ ] User-facing consent / disclosure shown BEFORE the session `url` opens (Didit handles biometric consent inside its flow but the legal layer outside is the integrator's responsibility).
- [ ] (Optional) `/v3/session/{id}/generate-pdf` surfaced in compliance UI.

## When you need more detail

- API root: `https://verification.didit.me/v3/` (everything `/v3/`). Auth only: `https://apx.didit.me/auth/v2/programmatic/register/`.
- Sessions overview: `https://docs.didit.me/sessions-api/overview`
- Standalone APIs index: `https://docs.didit.me/standalone-apis/id-verification`
- Programmatic registration: `https://docs.didit.me/integration/programmatic-registration`
- AI agent / MCP integration: `https://docs.didit.me/integration/ai-agent-integration`
- Webhooks reference: `https://docs.didit.me/integration/webhooks`
- SDKs overview: `https://docs.didit.me/integration/sdks`
- Full OpenAPI 3 spec: `https://docs.didit.me/openapi-25.json`
- MCP server config (drop into `.cursor/mcp.json`, `claude_desktop_config.json`, or equivalent):

```json
{
  "mcpServers": {
    "didit": {
      "command": "npx",
      "args": ["@didit-protocol/mcp-server"],
      "env": { "DIDIT_API_KEY": "your_api_key" }
    }
  }
}
```

The MCP server exposes 40+ tools spanning auth, sessions, workflows, questionnaires, users, billing, blocklist, and every standalone API — see `https://docs.didit.me/integration/ai-agent-integration` for the full tool catalogue.

Now build the integration. If anything in `## My application context` is missing, ask once at the top of your reply, then ship the complete change set: env vars wired, backend route, frontend SDK call, webhook endpoint with signature verification, decision-handling DB logic, and idempotent dedupe. Use my stack's idioms (`fetch` / `axios` / `requests` / `okhttp`), keep the API key server-side only, and follow the verification status state machine above.