The intelligence.
Your application.

Authentication

Create and manage keys above using a verified account. Only a prefix is available after creation. Rotation replaces a key and immediately revokes the previous one; revocation blocks further authentication. Internal service credentials are separate from public developer keys.

Endpoints and credits

These routes use one shared gateway and existing data services. Source configuration, read-only access, licensing approval and dataset coverage control availability. A listed route does not promise a connected source or a complete blockchain history.

Credits are reserved atomically before work starts. Failed data reads and work refused before execution are refunded. If research completion is uncertain, its credits remain reserved. Reusing the same request identifier cannot start another job. Check research history before creating a new request. Unavailable capabilities do not consume credits. Developer credits are separate from AskPumpGPT questions; existing Quant Research entitlement still applies to backtesting. Future API plans remain inactive.

Query parameters and resource limits

Collection requests accept limit (1–100) and offset. Every request is bounded by gateway and source timeouts, response limits, permissions and account limits. Use the returned pagination metadata to request the next page.

Request examples

# Run from your server. Store the key in a secret manager.
# Reuse this idempotency value only for retries of this exact request.
curl 'https://askpumpgpt.ai/v1/tokens?limit=20&offset=0' \
  -H "Authorization: Bearer $PUMP_API_KEY" \
  -H 'Idempotency-Key: token-page-2026-10-19-001'

curl 'https://askpumpgpt.ai/v1/research' \
  -X POST \
  -H "Authorization: Bearer $PUMP_API_KEY" \
  -H 'Content-Type: application/json' \
  -H 'Idempotency-Key: research-2026-10-19-001' \
  --data '{"question":"What does the available dataset show about recent token launches?"}'

// Server-side JavaScript. Never expose this key in a browser bundle.
const response = await fetch('https://askpumpgpt.ai/v1/analytics?hours=24&limit=24', {
  headers: {
    Authorization: 'Bearer ' + process.env.PUMP_API_KEY,
    'Idempotency-Key': requestId, // Keep stable when retrying this request.
  },
  signal: AbortSignal.timeout(30_000),
})
const result = await response.json()
if (!response.ok) throw new Error(result.error.message)
// Inspect result.meta.limitations and result.meta.pagination.

Responses, evidence and retries

Data routes return data and meta. Collection metadata includes pagination.limit, offset, nextOffset and hasMore. Metadata describes the dataset, data version, observation time and limitations. Research returns its report and evidence in the research contract documented in OpenAPI.

Provide a unique Idempotency-Key for each logical request and keep it unchanged when retrying the same method, path, parameters and body. Research requires this header. Reusing a key for different input causes a conflict. A completed successful request can replay without another credit charge. Failed, running or uncertain requests return 409 rather than starting another job.

Honor Retry-After for rate limits. If a response is lost, retry with the same identifier to avoid duplicate work. A 409 requires checking the existing request or research history; it does not resume the job. Inspect source limitations before comparing datasets. Available aggregates do not establish full historical coverage, and this API does not promise real-time data.

Error reference

Errors use { "error": { "code": "…", "message": "…" }, "requestId": "…" }. Keep the response request ID for support; exclude your API key from logs and error reports.