Evidence-grounded agent2026Public prototype

PropOps

An open-source research agent that joins government property records into an inspectable diligence workflow.

Back to projectsRepository

At a glance

Problem

Important property decisions are slowed down by fragmented records and hard-to-interpret legal signals.

Approach

Nineteen agent workflows combine registry, RERA, and court data with entity resolution, risk checks, and batch evaluation.

Why it matters

The output keeps sources, gaps, and judgment visible instead of hiding them behind one score.

System sketch

01Gather

Collect registration prices, project records, complaints, litigation, and listing context from public sources.

02Normalize

Resolve builder identities across legal entities, naming variants, contact details, directors, and addresses.

03Surface risk

Flag missing records, conflicting claims, delayed projects, litigation, contract clauses, and financial stress.

04Support judgment

Preserve the source trail and reserve outreach, negotiation, legal review, and purchase decisions for the buyer.

Design notes

Entity resolution changes the answer

A project-level lookup misses history when one builder operates through multiple legal entities and name variants.

Sources should remain inspectable

Every summary is more useful when a buyer can inspect the records, limitations, and unresolved gaps behind it.

Automation stops before the decision

The agent gathers, compares, and drafts. The human controls outreach, legal review, negotiation, and purchase.

Question

Indian property records are public, but the useful evidence sits across registration portals, state RERA systems, court databases, builder entities, and listing sites.

PropOps turns that fragmented search into a repeatable research workflow while preserving the source trail and known limitations.

Approach

  • Route a question through 19 modes covering discovery, evaluation, builder research, litigation, finance, agreement review, and post-purchase checks.
  • Cross-reference IGRS, state RERA portals, the MoHUA aggregator, and eCourts using dedicated scrapers and fallbacks.
  • Resolve related builder entities before aggregating project history, complaints, and litigation.
  • Run batch evaluations with parallel agents while leaving high-impact actions with the buyer.

Open questions

  • How can scraper failures and portal changes be detected before they create false confidence?
  • Which evidence model makes missing and contradictory records easy to review?
  • How far can entity resolution go before a probable match needs explicit human confirmation?

Project spec

RoleProblem framing, systems design, and implementation.
StatusPublic prototype
TypeEvidence-grounded agent
UpdatedApr 2026
Primary objectAn evidence-linked property or builder report with explicit risks and open questions.
System boundaryIGRS, five state RERA portals, a national aggregator, and eCourts.
Current artifactTypeScript and Playwright research workflows, plus a Go terminal dashboard.
Pressure pointPortal drift, incomplete disclosures, and cross-entity builder matching.

Topics

agent workflowsgovernment dataentity resolutionhuman-in-the-loop

Where it is now

  • PropOps is open source and includes one-command installers for non-technical users.
  • The builder pipeline combines state-specific scrapers, cross-entity matching, and national court search.
  • The system never contacts builders or submits applications; the buyer retains control of consequential actions.