Task API
Automate structured web prospecting & deep research
Transform manual knowledge workflows into programmable, repeatable operations. Combines state-of-the-art web search, live multi-page crawling, and reasoning agents.
Parallel Task API
from parallel import Parallel
client = Parallel(api_key="PARALLEL_API_KEY")
# Create an autonomous web research task
task = client.task.create(
objective="Analyze enterprise AI agent adoption in Fortune 500 financial institutions",
input_schema={
"institutions": ["JPMorgan", "Goldman Sachs", "Morgan Stanley", "Citi"]
},
output_schema={
"deployments": [
{
"institution": "string",
"use_case": "string",
"vendor_or_inhouse": "string",
"reported_roi": "string",
"source_citations": ["string"]
}
]
},
processor="ultra" # lite | core | pro | ultra
)
# Stream progress events and live reasoning trace
for event in task.stream():
print(f"[{event.stage}] {event.message} (Sources evaluated: {event.sources_count})")
result = task.wait()
print(result.data)Efficiency
Save human hours with structured web search tasks
Turn complex knowledge work that previously took human analysts weeks into repeatable workflows that take just minutes.
100x
Faster research cycle
99.2%
Fact grounding rate
Autonomous Execution Flow
1Decompose objective into 15+ sub-queries across search dimensions
2Crawl & parse authoritative primary source domains and PDFs
3Cross-verify facts, resolve conflicting statements, and generate Basis graph
✓Return structured JSON & cited markdown report directly to your agent
Extensibility
Build web agents with unrivaled flexibility
Create task specifications for any research need: market intelligence, due diligence, lead lists, or compliance monitoring.
Objective Prompt
"Research the current market landscape for open-source agent frameworks in 2026. Compare runtime architectures, benchmark scores on GAIA, and license terms."
Output JSON Schema
{
"frameworks": [
{
"name": "string",
"architecture": "hierarchical | swarm | react",
"gaia_score": "number",
"license": "string",
"key_differentiator": "string"
}
],
"market_trends_summary": "string"
}Trust & Provenance
Verifiability and provenance for every atomic fact
Every output includes Parallel's Basis framework—a proprietary grounding graph linking every generated sentence directly to timestamped source excerpts and confidence scores.
- Exact paragraph coordinates on original HTML pages
- Cross-source corroboration across multiple independent domains
- Calibrated uncertainty quantification (0.0 to 1.0 confidence)
Basis Provenance Example
"Parallel ranks #1 on the Artificial Analysis Search Index for AI agent retrieval quality."
Source: artificialanalysis.ai/agents/search-api (Confidence: 0.99)
FAQ
No, input schema is completely optional. You can simply pass a plain natural language prompt like 'Analyze semiconductor supply chain constraints in 2026' or provide a structured schema for rigid database extraction.