generated: '2026-08-09'
method: derived
source: >-
  openapi/connext-everclear-openapi.yml, https://docs.everclear.org/developers/api.md,
  https://docs.everclear.org/developers/guides/quickstarter.md
summary: >-
  Cross-cutting runtime semantics for the Everclear ("Chimera") REST API, derived from the 17
  OpenAPI documents Everclear publishes. The API is a thin read/build layer over on-chain
  protocol state: GETs read indexed intents, invoices, assets and history; POSTs do not
  mutate anything server-side, they return an unsigned TransactionRequest for the caller to
  sign and broadcast. That shape is why several conventions a normal write API needs
  (idempotency keys, auth, rate-limit headers) are simply absent.
authentication:
  style: none
  detail: No securityScheme is declared and no operation carries a security requirement.
  see: authentication/connext-authentication.yml
idempotency:
  supported: false
  detail: >-
    No Idempotency-Key header, no idempotency parameter, no request-fingerprint field, and no
    mention of idempotency in the docs. It is not needed in the usual sense: no POST in this
    API commits state on Everclear's side — POST /intents, POST /solana/intents,
    POST /tron/intents, POST /intents/{intentId}/execute and
    POST /intents/{intentId}/return-unsupported are all transaction builders that return
    calldata. Replay protection lives on-chain in the EVM nonce and in the Permit2 nonce
    supplied by the caller in IntentRequest.permit2Params.
  on_chain_replay_protection:
    - TransactionRequest.nonce (account nonce, chosen by the signer)
    - IntentRequest.permit2Params.nonce + deadline + signature (Permit2 approval replay guard)
    - IntentRequest.order_id (caller-supplied order identifier for newOrder requests)
pagination:
  primary_style: cursor
  detail: >-
    Collection reads take cursor, prevCursor and limit as query parameters and return
    nextCursor, prevCursor and maxCount alongside the item array. Cursors are opaque strings
    and are nullable at the ends of the range.
  request_params: [cursor, prevCursor, limit]
  response_fields: [nextCursor, prevCursor, maxCount]
  default_limit: 10
  operations:
    - GET /intents
    - GET /batched-intents
    - GET /invoices
    - GET /history/{intentId}/invoice-processing
  inconsistency:
    operation: GET /configs/assets
    style: offset
    params: [page, limit]
    note: >-
      The only paginated endpoint that uses page/limit instead of cursors — and both are
      REQUIRED there, unlike every cursor endpoint where they are optional. It also returns
      no nextCursor/maxCount, so a client cannot tell when it has reached the end.
filtering:
  detail: >-
    Intent and invoice collections filter on repeated query arrays and scalars rather than a
    query language.
  parameters:
    - statuses (array of IntentStatus)
    - origins / destinations / allDestinations (arrays of domain ids)
    - txHash, userAddress, tickerHash
    - startDate / endDate (string)
    - isFastPath (boolean — intents with TTL > 0)
    - sortOrderByDiscount (invoices only)
identifiers:
  intent_id: 32-byte hex intent id, used as the path parameter on every /intents/{intentId} and /invoices/{intentId} route
  batch_id: batch identifier returned on /batched-intents
  ticker_hash: hash identifying a token across domains
  domain: >-
    Everclear addresses chains by "domain" — a numeric string identifier (mostly the chain id)
    passed as origin/destinations throughout.
  canonical_id / canonical_domain: canonical asset identity used by the Asset object
numbers_and_money:
  detail: >-
    All amounts are strings, not numbers, with format bigint, denominated in the asset's own
    decimal units on the origin chain. Fees come back as fixedFeeUnits (string/float),
    variableFeeBps and totalFeeBps (number). Asset decimals are exposed as decimal and
    adopted_decimal on the Asset object.
versioning:
  scheme: unversioned
  detail: No version segment in the path and no version header. See lifecycle/connext-lifecycle.yml.
error_envelope:
  media_type: application/json
  shape: '{error: string, message: string}'
  rfc9457: false
  see: errors/connext-problem-types.yml
rate_limits:
  documented: false
  detail: >-
    No rate-limit policy is published, no 429 response is declared on any operation, and no
    RateLimit/Retry-After headers are documented.
request_tracing:
  supported: false
  detail: No request-id / correlation-id header is documented or returned in the error envelope.
webhooks_or_events:
  supported: false
  detail: >-
    No webhook or callback surface. Event consumption was intended to happen through the
    Goldsky subgraphs (see graphql/connext-everclear-subgraph.graphql) and, for fill agents,
    an Envio GraphQL indexer — every documented subgraph URL now returns 404.
field_expansion:
  supported: false
sparse_fieldsets:
  supported: false
metadata_fields:
  supported: false
cross_links:
  errors: errors/connext-problem-types.yml
  lifecycle: lifecycle/connext-lifecycle.yml
  authentication: authentication/connext-authentication.yml
  data_model: data-model/connext-data-model.yml
caveat: >-
  Derived from the published contract only; https://api.everclear.org returned HTTP 530 on
  2026-08-09, so no live response headers could be observed.