AI that can show you why it knows.
VeritasGraph combines knowledge-graph reasoning, hierarchical tree search, retrieval and verifiable provenance to turn enterprise information into grounded, auditable AI answers — with governed agents on your infrastructure.
Watch a question
become evidence.
Run a simulated GraphRAG request and watch semantic retrieval, graph traversal, multi-hop reasoning and provenance resolve in sequence.
The incident is associated with a gateway timeout introduced during deployment v4.8 in the APAC environment.
Structure + semantics
beats similarity alone.
VeritasGraph can combine graph relationships, hierarchical document context and retrieval signals instead of treating every chunk as an isolated island.
Traditional RAG
GraphRAG
Navigate hierarchy.
Reason across relationships.
A document tree answers “where is the relevant context?” while a graph answers “what is connected to what?” The two views can cooperate during retrieval.
Every important claim
has a trail.
Instead of presenting citations as decoration, the interface makes provenance a first-class object: claim → source span → graph relationship → evidence.
AI that is
allowed to act.
Studio-style orchestration can place guardrails, memory, knowledge, context budgeting, tools and data logging around every agent turn.
Run a governed turn.
The simulation shows how an agent request can pass through explicit control boundaries before it reaches tools or produces an auditable response.
awaiting request...
Guardrails are
part of the path.
Show a recruiter that governance is not an afterthought: policy checks, PII redaction and audit events can be visible in the same execution trace.
Block unsafe input before action.
Click the simulation to watch a sensitive payload get detected, redacted or blocked before the agent reaches a tool.
no active events
Make decisions inspectable.
From citizen report
to verified case.
A concrete workflow from the repository: incident reporting, computer-vision validation, knowledge-graph routing, evidence fusion and case registration.
Overflowing waste near market
Citizen report · photo attached · location resolved · graph context available
Evidence fusion, not guesswork.
Connect the graph
to agent ecosystems.
VeritasGraph includes an MCP bridge pattern so compatible clients can discover and invoke graph capabilities through explicit tools.
Agent surfaces
MCP ServerJSON-RPC · tool discovery · graph access
Explicit tools
Under the hood.
A recruiter-friendly view of how the system moves from raw information to an answer that can be traced, governed and inspected.
Documents
PDF, text, structured or application data.
Extraction
Entities, relations, sections and evidence.
Graph Store
Nodes, typed edges and grounded source links.
Hybrid Search
Graph + tree + semantic retrieval signals.
Multi-hop Context Engine
Find entry points, traverse relevant relationships, fuse hierarchy and assemble evidence-aware context for the model.
Guardrails
Policy checks, redaction and governance boundaries.
Agents + Tools
Memory, tools, MCP and governed orchestration.
Verified answer + provenance + audit trace
The interface exposes the path so users can understand not only what the system answered, but which entities, relationships and source spans supported it.
Built as an
engineering system.
Technology grouped by responsibility so a recruiter can scan the engineering surface without wading through a giant logo wall.
Query playground.
What caused the APAC payment incident?
Run a query to see a simulated grounded answer with an inspectable reasoning trace.