AI investigation workspace — live agents

Turn a question into a quest.

QuestAI investigates your goal using specialized AI agents, real-world search, and structured reasoning — then gives you an answer you can actually use. Not a chatbot. A workspace.

LangGraph orchestrationGroq gpt-oss-120B / 20B · Qwen 27BSerpApi Google · Flights · Hotels · Shopping · NewsFastAPI + Pydantic · Vite React
4 Specialized agents·6 SerpApi engines·Evidence preserved·Structured results
What do you want to accomplish?Research mode
Give QuestAI a goal to investigate.
⌘ + Enter
GoalInvestigationIntelligenceResult
Agent system — evaluated from code

One goal. Specialized intelligence.

QuestAI routes your goal to the right LangGraph. Each agent has its own tools, schemas, and evidence pipeline — not a prompt wrapper.

Agent 01 · Research · Live

Research Agent

Evidence-grounded research with citation validation. LLM proposes → evidence preserved → validator checks.

graph: planner → router → collector → evaluator → gap_analyzer → synthesizer → validator
3 rounds · 15 evidence max · 1 resynthesis
tools: google / scholar / news via SerpApi
Agent 02 · Travel · Live

Travel Agent

Trip Planner orchestrates Flight + Hotel + destination research with dependency-aware execution.

trip_graph: 7 nodes · DEPENDENCY_MAP
hash-cached · conditional dispatch
Google Flights & Hotels via SerpApi
Agent 03 · Shopping · Live

Shopping Agent

PriceScope — live Indian Google Shopping comparison. Verbatim product evidence, cheapest highlight.

engine: google_shopping · INR · India
fast-path LLM cleaning (gpt-oss-20B)
evidence: verbatim store/price/source
Agent 04 · Jobs · Roadmap

Job Agent

Investigates opportunities by skills & location — structures matches with evidence. Demo only.

status: frontend demo · no backend yet
planned: google jobs via SerpApi
demo shows filtering & match scores
Conditional execution — real logic

The right agent.
For the right quest.

Travel's DEPENDENCY_MAP + input-hash caching means "change hotel to 5-star" only reruns Hotel Agent.

“Compare iPhone 16 prices” → Shopping Agent ✓

Research —   Travel —   Jobs — bypassed · saved 3 tool calls

trip_graph.py:32 DEPENDENCY_MAP · trip_graph.py:613 dispatch_agents · schemas/trip.py:152 compute_flight/hash
Your Goal
QuestAI Router
Research —
Travel —
Shopping
Jobs —
Trust — code-level

AI reasoning, grounded in evidence.

Evidence is extracted verbatim (ev_001), capped at 15, deduped by claim — then validator does atomic claim checks and can force one resynthesis.

MAX_SYNTHESIS_EVIDENCE=15 · max_rounds=3 · resynthesis_count<1

services/source_processor.py · graph/nodes/validator.py:213 · hedged findings require evidence_refs
Sources
Evidence
ev_001
Analysis
Validation
Answer
How it works

From intent to useful output.

A focused workflow replaces the blank chat box and the wall of generated text.

01

Give it a goal

Describe what you want — trip, product, research question — in natural language.

02

QuestAI investigates

Router picks the right LangGraph. Dependency-aware dispatch runs only needed tools.

03

Evidence becomes insight

Verbatim evidence is deduped, hedged, and validated atomically before synthesis.

04

Get something useful

Structured cards, comparisons, itineraries — not a chat transcript.

Product philosophy

Not another chatbot.

QuestAI is built around goals, investigation, specialized agents, and structured results — not endless conversations.

Give QuestAI a goal. It investigates. You get the result.

SerpApi India Hackathon 2026

Why we built QuestAI

Search results alone are not answers. We built QuestAI to turn a goal into an investigation — specialized agents use SerpApi live data to plan, collect, validate and synthesize, delivering structured workspaces instead of chat dumps.

serpapi.github.io/serpapi-india-hackathon-2026
Your next quest

What will you investigate?

Give QuestAI your next question, decision, trip, or product search.

Demo — frontend only