The graph beneath the web

Give your agents answers that search can't return.

Lokaah turns the web into a connected graph of entities and relationships, so your agents can answer questions that span multiple sources.

THE SEARCH BOTTLENECK

Search finds pages. Your agent still has to connect them.

Web search dumps unverified HTML for humans to browse. Lokaah provides a pre-resolved, traversable entity graph in a single MCP call.

The Web Search Loop
Fuzzy String Matching
Fragile
Architecture · Entity ResolutionPre-resolved canonical IDs
Unresolved Web (Fuzzy)Companies House #08SEC CIK#000182MCA L65990MH"Peak Partners"peak-ventures-ukResolveCanonical Graph (Resolved)Canonical EntityPeak Venture Partnersid: ent_pk_902 (Global)UK Cos HouseSEC EDGARMCA India
10-20 disparate candidate URLs with homonym ambiguity
8,400 tokens per question wasted on HTML boilerplate
15 - 30s latency with unverified hallucinated links
The Lokaah Graph
Deterministic Token Efficiency
1 MCP Call
Context Load · Token Tape-61% Context Overhead
Web Search (3 SERPs + 2 PDFs)8,400 tokens
HTML BoilerplateNav Chrome & AdsActual Fact Signal
Lokaah MCP (1 Typed Graph Payload)3,300 tokens (Pure Signal)
Pre-resolved canonical IDs across international corporate registries
3,300 tokens of pure structured signal (-61% context load)
< 90ms latency with cryptographic statutory receipts
WEB SEARCH VS LOKAAH MCP

What happens when an agent asks a question.

Pick a real question below. See how standard web search gets lost in unstructured pages, while Lokaah returns an exact, statutory answer in a single MCP tool call.

User Prompt to AI Agent Live Question
"Does Wipro share any active board directors with other Indian financial institutions?"
Approach A

With Web Search Only

Unstructured HTML
Agent scrapes multiple search results:
SERP #1:site:wikipedia.org "Wipro board of directors"
SERP #2:"Deepak Satwalekar" directorships financial companies
SERP #3:"Wipro" "Home First" shared director news
Agent Output (Fragmented)

Wipro lists several independent directors. Articles mention Deepak Satwalekar in multiple contexts, but search cannot cross-reference Ministry of Corporate Affairs DIN records to confirm active overlapping seats without manually checking annual PDF filings.

8,400 tokens14.2s12 unverified web pages
Uncertain link · High token cost · No statutory proof
Approach B

With Lokaah MCP

1 Typed Tool Call
Single Deterministic MCP Tool Call:
lokaah_finance_entity_neighbors({ symbol: "WIPRO", edge_type: "director_interlock" })
Traversed Graph NodesIndependent Director interlock
Wipro Limited
NSE: WIPRO
Deepak Satwalekar
MCA DIN: 00009627
Home First Finance
NSE: HOMEFIRST
Verified Answer

Deepak Madhav Satwalekar (DIN 00009627) sits on the boards of both Wipro Limited and Home First Finance Company India Limited as an Independent Director.

94 tokens48msMCA India Form DIR-12
Proof:MCA DIN 00009627 · MCA CIN L32102KA1945PLC020800 & L65990MH2010PLC240703
FREQUENTLY ASKED QUESTIONS

Everything you need to know about the graph layer.

Direct answers on architecture, entity coverage, and how Lokaah differs from general web scraping.

01

Is this a search API?

No. Search engines return unstructured documents and HTML pages for humans to read. Lokaah returns a pre-resolved, traversable graph of entities, directional relationships, and statutory citations in a single typed MCP call.

02

Why not just ask an LLM with web search?

A model with web search has to scrape pages, parse raw HTML, and guess connections at runtime. It consumes thousands of tokens on boilerplate, hallucinates links, and cannot guarantee regulatory accuracy. Lokaah provides a deterministic graph where every edge carries its primary statutory source.

03

Where does the data come from?

Official public registries, regulatory filings, corporate gazettes, and statutory records across global jurisdictions (including UK Companies House, SEC EDGAR, MCA India, and exchange disclosures). Every fact is anchored to its original source document and timestamp.

04

Does this index our internal company files?

No. Lokaah is built exclusively for the public, statutory, and regulatory web. We do not index internal documents, Notion workspaces, or private company databases.

05

How does an agent integrate with Lokaah?

Through standard Model Context Protocol (MCP) or our REST API. Any agent running in Claude, Cursor, OpenAI, or custom multi-agent frameworks can call Lokaah tools directly to resolve entities, walk multi-hop relationships, and verify citations.

06

How is pricing and beta access structured?

We are currently rolling out in beta with selected design partners. Access is enabled per workspace with dedicated support and agreed terms for production volume.

Let your agent see what the web does not show.