Async multi-hop web research tasks via Parallel, paid per-call with pathUSD on Tempo.
What it does
The Parallel Task endpoint (`POST /api/task`) runs asynchronous, multi-hop web research tasks powered by Parallel Web Systems. You submit a research query along with a processor tier — "ultra" ($0.30/task, highest quality) or "pro" ($0.10/task, faster) — and receive a `run_id`. You then poll `GET /api/task/{runId}` (free, no payment required) until the task status reaches "completed", which can take 1–5+ minutes depending on complexity and tier.
Payment is handled via the Machine Payment Protocol (MPP) on the Tempo blockchain using pathUSD. The endpoint returns an HTTP 402 challenge; the `npx mppx` CLI or SDK signs the challenge and retries automatically. No manual credential construction is needed. Rate limits are 60 requests per minute per IP, and all payments are final with no refunds once confirmed on-chain.
This endpoint is part of the broader Parallel API gateway, which also offers web search (`/api/search`, $0.01/query) and URL content extraction (`/api/extract`, $0.01/URL). Free utility endpoints include task polling and wallet balance checks. The full API schema is available at `GET /api` for programmatic self-discovery by agents.
Capabilities
Use cases
- —Running deep, multi-step web research on a topic (e.g., market overviews, competitive analysis) without manual browsing
- —AI agents autonomously performing paid research tasks and polling for results
- —Generating comprehensive research briefs from a single natural-language query
- —Combining search and extraction in a single async pipeline for structured intelligence gathering
Fit
Best for
- —AI agents needing autonomous deep web research capabilities
- —Developers building research automation pipelines with pay-per-use pricing
- —Tasks requiring multi-hop reasoning across multiple web sources
- —Scenarios where comprehensive quality matters more than sub-second latency
Not for
- —Real-time, low-latency lookups (tasks take 1–5+ minutes; use /api/search with mode "fast" instead)
- —Free or unpaid usage — every task call requires pathUSD on Tempo
- —Use cases requiring refunds or reversible transactions (all payments are final)
Quick start
# 1. Create a Tempo account
npx mppx account create
# 2. Fund wallet with pathUSD, then run a task
npx mppx https://parallelmpp.dev/api/task --method POST -J '{"input":"HVAC market overview USA","processor":"ultra"}'
# 3. Poll for results (free)
curl https://parallelmpp.dev/api/task/<run_id>Example
Request
{
"input": "Brief HVAC market overview USA",
"processor": "ultra"
}Response
{
"run_id": "trun_xxxxx",
"status": "pending",
"poll_url": "/api/task/trun_xxxxx"
}Endpoint
Quality
Full OpenAPI schema with clear pricing, request/response structure, and detailed agent integration docs. The probe did not capture a live 402 challenge (POST-only endpoint was probed with HEAD/GET, returning 404), but the OpenAPI spec and crawled docs are comprehensive. Response schema for completed tasks is not documented, so the example response is inferred from the docs' description of run_id and poll_url.
Warnings
- —Probe returned 404 because the endpoint only accepts POST; this does not indicate the endpoint is down.
- —No detailed response schema for completed task results is documented — only run_id/poll_url on creation and status on polling.
- —All payments are final; no refunds once confirmed on-chain.
- —The /docs and /pricing pages return 404; documentation lives on the homepage and GET /api.
Citations
- —Task endpoint pricing: $0.30 for ultra, $0.10 for pro processorhttps://parallelmpp.dev/api
- —Tasks are asynchronous and can take 1–5+ minutes; poll GET /api/task/{runId} until completedhttps://parallelmpp.dev
- —Payment via MPP/Tempo using pathUSD; 402 challenge handled by npx mppxhttps://parallelmpp.dev
- —Rate limit: 60 requests/minute per IPhttps://parallelmpp.dev
- —All payments are final, no refunds available once confirmed on-chainhttps://parallelmpp.dev/api
- —Polling GET /api/task/{runId} is free — no payment requiredhttps://parallelmpp.dev