---
title: "GraphQL endpoint"
method: POST
path: "/v3/graphql"
tags: ["GraphQL"]
---

# GraphQL endpoint

`POST /v3/graphql`

GraphQL endpoint (shared auth/rate limit with REST), proxied to the
Ponder GraphQL API.
Ponder LIKE filter operators reject unescaped `%` and `_` wildcard values.
Before proxying, the API bounds the complete request including unused runtime variables,
variable defaults, inline arguments, input bytes/nodes/depth, boolean clauses, list entries,
aliases, `totalCount`, execution cost, `limit` (default maximum 1,000), and `offset`
(default maximum 10,000). Offset scan cost is `offset + limit`; selected output expansion
remains based on `limit`. Resource violations return `QUERY_TOO_BROAD` and malformed
documents return `VALIDATION_ERROR` without contacting Ponder. Ponder 0.15.17 does not use
`first` or `last` pagination arguments.
Each API process also enforces a shared upstream concurrency cap (default 8), one total
fetch-and-stream timeout (default 10 seconds), redirect rejection, GraphQL JSON media-type
and object-shape validation, and a streamed decoded response cap (default 5,242,880 bytes).
A non-queueing per-process query-cost bucket defaults to 100,000 units of capacity and a
25,000-unit-per-second refill. Exhaustion returns 429 before an upstream slot is acquired;
the shared per-client API rate limiter remains the outer control.

## Request body

- object
  - `query` string
  - `variables` object
  - `operationName` string

## Response `200`

GraphQL response

- object

## Other responses

- `400` — Bad request
- `413` — Request body too large
- `415` — Unsupported media type
- `429` — Rate limited
- `502` — Upstream dependency failure
- `503` — Service unavailable

---

[API](https://skmtc.dev/euler/apis/euler-data-api-v3.md) · [All operations](https://skmtc.dev/euler/apis/euler-data-api-v3/llms.txt) · [OpenAPI document](https://skmtc-service-production.skmtc.workers.dev/v1/apis/euler/euler-data-api-v3/revisions/9573a5fbd4a6/schema)