> ## Documentation Index > Fetch the complete documentation index at: https://docs.trycherry.ai/llms.txt > Use this file to discover all available pages before exploring further. # Upload a bank statement (write scope) > Send a statement file and Cherry does the rest: transcribes it, verifies the transactions roll the opening balance forward to the printed closing balance (a statement that doesn't reconcile is rejected whole), creates or matches the manual institution + account, and lands the transactions — deduplicated against rows pushed via POST /transactions. Processing is asynchronous: the response is a `202` with a `documentId`; poll GET /bank-statements/{documentId} for the outcome (typically well under a minute). Re-uploading the same statement is a no-op: the period is keyed by (account, periodStart, periodEnd) and each transaction by its content, so nothing duplicates. PDF is the best input; images of statements work too. 25 MB max. Requires the `write` scope. Rate-limited more tightly than other writes — each upload costs a transcription pass. ## OpenAPI ````yaml /openapi.json post /bank-statements openapi: 3.0.3 info: title: Cherry API version: 1.0.0 description: >- Cherry's public developer API — programmatic access to a business's books as Cherry keeps them: bank transactions with their categorizations, the chart of accounts, the double-entry journal, financial statements, legal entities, the tax-obligation calendar, and the fiscal health score. Authenticate every request with an API key: `Authorization: Bearer ck_live_…`. Keys are created in the Cherry app and are scoped to one business (tenant) — there is no tenant id in any URL. Keys carry a `read` and/or `write` scope; every GET needs `read`, mutations need `write`. Conventions: successful responses wrap the payload in `{ "data": … }`; list endpoints add `{ "meta": { "pagination": { limit, offset, nextOffset } } }` — pass `nextOffset` back as `offset` until it is null. Errors are `{ "error": string, "reason"?: string }` with conventional status codes. Dates are `YYYY-MM-DD`; money amounts are numbers in the row's currency. Requests are rate-limited per key with token buckets (429 + `Retry-After` when exhausted); report generation has a smaller budget than plain reads. servers: - url: https://api.trycherry.ai/v1 security: - bearerAuth: [] tags: - name: Transactions description: Bank feed rows and their categorizations. - name: Accounts description: The tenant's chart of accounts. - name: Bank accounts description: >- Connected bank accounts — Plaid-linked and manual. Manual institutions are how developers without a Plaid connection bring their own accounts and transactions. - name: Bank statements description: >- Statement uploads — send a statement PDF/image and Cherry transcribes it, verifies the balances reconcile, and lands the transactions on a manual bank account. The no-export alternative to pushing rows yourself. - name: Journal description: The double-entry journal (entries and lines). - name: Reports description: Trial balance and financial statements. - name: Entities description: Legal entities behind the tenant. - name: Obligations description: The tax/compliance calendar Cherry tracks. - name: Health description: Cherry's composite fiscal readiness score. - name: Meta description: The API's own machine-readable contract. paths: /bank-statements: post: tags: - Bank statements summary: Upload a bank statement (write scope) description: >- Send a statement file and Cherry does the rest: transcribes it, verifies the transactions roll the opening balance forward to the printed closing balance (a statement that doesn't reconcile is rejected whole), creates or matches the manual institution + account, and lands the transactions — deduplicated against rows pushed via POST /transactions. Processing is asynchronous: the response is a `202` with a `documentId`; poll GET /bank-statements/{documentId} for the outcome (typically well under a minute). Re-uploading the same statement is a no-op: the period is keyed by (account, periodStart, periodEnd) and each transaction by its content, so nothing duplicates. PDF is the best input; images of statements work too. 25 MB max. Requires the `write` scope. Rate-limited more tightly than other writes — each upload costs a transcription pass. operationId: uploadBankStatement requestBody: required: true content: multipart/form-data: schema: type: object required: - file properties: file: type: string format: binary description: The statement file (PDF or image). Max 25 MB. entityId: type: string format: uuid description: >- Legal entity the account belongs to (UUID from GET /entities). Defaults to the tenant's primary entity. responses: '202': description: Accepted — processing has started. content: application/json: schema: type: object properties: data: type: object properties: documentId: type: string format: uuid status: type: string enum: - processing example: data: documentId: 9f8e7d6c-5b4a-4321-8765-4321fedcba98 status: processing '400': description: Error content: application/json: schema: $ref: '#/components/schemas/Error' '401': description: Error content: application/json: schema: $ref: '#/components/schemas/Error' '403': description: Error content: application/json: schema: $ref: '#/components/schemas/Error' '413': description: Error content: application/json: schema: $ref: '#/components/schemas/Error' '429': description: Error content: application/json: schema: $ref: '#/components/schemas/Error' components: schemas: Error: type: object required: - error properties: error: type: string description: Stable machine-readable error code. reason: type: string description: Human-readable explanation. securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: ck_live_… description: >- Cherry API key, created in the app. Sent as `Authorization: Bearer ck_live_…`. ````