Schemes are fragmented
Benefits are spread across central and state portals, notices, PDFs, and department pages.
Citizen-first welfare discovery
Find schemes you may likely qualify for, with sources and proactive alerts.
SchemeSeva is a civic AI agent that helps citizens discover relevant government schemes, understand likely eligibility, and keep watch for future matches.
No sign-up required for the demo. Reports are guidance, not government approval.
What happens after getting started?
The agent builds a profile, searches schemes, checks rules, and validates the report.
- Data needed
- State, age, category, occupation, income, document status
- Not collected
- Aadhaar numbers, bank account numbers, passwords, or uploads
- Output
- Likely eligible schemes with sourceUrl and lastVerified
The problem
Welfare discovery should not depend on luck.
Eligibility is confusing
Rules change by state, category, age, occupation, income, documents, and special status.
Portals are mostly reactive
Citizens often have to know what to search for before they can discover a useful benefit.
People miss support
Farmers, students, elders, entrepreneurs, and field workers can lose time chasing the wrong path.
The agent
SchemeSeva puts the agent workflow at the center.
This is not just a chatbot. The app runs a structured workflow that gathers profile facts, retrieves schemes, remembers context, checks rules, writes a simple report, validates output, and lets Vigilance act on new matches.
Profile Agent
Structures basic profile details from a guided form or plain-language note.
Discovery Agent
Searches the verified scheme catalog with Qdrant retrieval or honest local fallback.
Eligibility Agent
Checks rules such as state, age, occupation, income, category, and documents.
Report Agent
Explains likely matches, documents, next steps, sources, and last verified dates.
Vigilance Agent
Keeps watch after the first search and can raise proactive alerts for new matches.
Enkrypt validation
Checks citizen-facing outputs before they are shown.
Workflow
From profile to source-grounded report.
- 1
Share basic profile details.
- 2
The agent searches verified Central + Telangana schemes.
- 3
Rules check likely eligibility.
- 4
The report shows schemes, documents, next steps, sources, and last verified dates.
- 5
Vigilance can alert when a new matching scheme appears.
Capabilities
Built for useful, inspectable recommendations.
Guided profile form
A clear intake flow that keeps required fields visible.
28 verified schemes
Focused Central + Telangana catalog, without overclaiming national coverage.
Qdrant semantic retrieval
Finds relevant schemes beyond exact keyword matching.
Featherless reasoning
Uses open-source model reasoning for reports and alert reasons.
Enkrypt safety validation
Validates reports and Vigilance alerts before display.
Proactive Vigilance alerts
Demonstrates watch-after-search behavior for matching schemes.
Source-grounded reports
Every result keeps sourceUrl and lastVerified visible.
Privacy-conscious handling
The demo uses a session key and does not collect Aadhaar numbers or bank account numbers.
Transparent debug page
Judges can inspect integration status without exposing secrets.
Benefits
Useful for citizens and the people helping them.
Saves time during scheme discovery.
Reduces confusion around rules and documents.
Finds relevant schemes faster than manual browsing.
Shows documents and next steps in simple language.
Keeps official sources close to every recommendation.
Keeps watching after the first search.
Helps citizens, students, farmers, entrepreneurs, elderly citizens, and field workers.
Use cases
One workflow, many citizen situations.
Farmer
Find farm, irrigation, income support, and equipment schemes they may likely qualify for.
Student
Discover scholarships, fee support, and education-related benefits with source links.
Woman entrepreneur
Spot enterprise, livelihood, and women-focused support programs faster.
Elderly pensioner
Check likely pension and social assistance matches with simple next steps.
Unemployed youth
Find skilling, employment, and self-employment schemes without portal hopping.
NGO / field worker
Use a repeatable workflow to guide citizens and document scheme sources.
Trust and safety
Honest guidance, not false certainty.
Guidance only
Reports use likely eligible language. Citizens should confirm details on official portals.
Validated output
Enkrypt checks reports and Vigilance alerts before citizen-facing display.
Privacy-conscious demo
No Aadhaar numbers or bank account numbers are collected. Source links and last verified dates stay visible.
Hackathon proof
The stack is visible, integrated, and judge-friendly.
SchemeSeva keeps technical proof close without turning the product into a dashboard. The debug page shows integration status while preserving fallbacks and avoiding secret exposure.
Mastra-style workflow
Typed multi-agent orchestration
Qdrant
Vector retrieval, session memory, and pending alerts
Featherless AI
Primary open-source reasoning for reports and alerts
Enkrypt AI
Report and alert validation
OpenRouter
Fallback reasoning provider
Gemini
Embeddings for semantic search
Langfuse
Observability traces
Upstash Redis
Rate limiting
Vercel
Deployed hackathon demo
Start with a demo profile or enter your own basic details.
Launch the agent, browse the verified catalog, inspect the architecture, or check live integrations for the hackathon proof trail.