Vedi and Vetra are Vetrn’s two agentic analysis engines. Each one deploys a coordinated swarm of specialised agents — orchestrated automatically — to conduct deep, structured research on a company and its sector simultaneously.
When you trigger a Vedi or Vetra run, a master orchestrator agent takes control. It understands the full context of the deal — the thesis, the deck, everything already in the deal’s memory layer — and then deploys multiple specialised agents, each assigned a specific research role.
These agents work simultaneously across their respective domains: one validates the founding team’s background, another maps the competitive landscape, another analyses regulatory risk, another builds the unit economics model. Each brings a distinct skill set. The orchestrator manages sequencing, handles dependencies between agents, and synthesises all findings into a single, structured, IC-grade output.
The entire process is fully automated — from the moment you trigger a run to the moment the report is ready.
Vedi analyses the company. Vetra analyses the sector. They run independently — so your view of the market is never shaped by the founder’s narrative.
Vedi conducts a deep, multi-dimensional analysis of the startup itself. It goes far beyond what’s in the pitch deck — validating claims, uncovering risks, and building a structured picture of the company across every dimension that matters at Seed to Series B.
Every finding is verified. Unverified claims are explicitly flagged. The output is structured and citation-backed — not an AI summary, but an institutional-grade company profile ready for IC use.
Vetra conducts independent sector research — completely separate from Vedi’s company analysis. This independence is intentional. When the same system analyses both company and market simultaneously, the company’s framing can bias the market view. Vetra removes that risk.
The result is an objective, founder-independent read on the sector: where the market is heading, what the competitive density looks like, what the exit environment supports, and where the regulatory and capital flow winds are blowing.
Most AI tools analyse a company and its market in the same pass — using the founder’s deck as the primary source for both. This means the market analysis is inevitably shaped by how the founder framed it. Vetra runs completely independently: no access to Vedi’s findings, no exposure to founder framing. You get a clean, objective read on the sector, then compare it to the company’s claims. That comparison is where real analytical value lives.
A single AI model asked to research everything produces mediocre results on everything. Specialised agents with defined roles produce expert-level outputs in their domains.
Each agent is designed for a specific research task — competitive analysis, legal signal extraction, financial modelling, and so on. Specialisation produces depth that generalist models can’t match.
Claims produced by research agents are passed through dedicated verifier agents before reaching the output. Unverifiable claims are explicitly flagged rather than stated as fact. This is structural, not optional.
The orchestrator understands dependencies between agent outputs — some agents can run in parallel, others need upstream results first. It sequences work intelligently, not arbitrarily.
The orchestrator synthesises all agent findings into a single structured output — formatted like an institutional research report, not a wall of AI text. Every section has a defined purpose.
Every agent run is informed by the full Vetrn Memory layer — everything already known about the deal informs what agents focus on and what they flag as inconsistent.
Which agents run, their priorities, and their research depth can be configured per fund. Your focus areas — whether that’s deep regulatory analysis or heavy emphasis on founder track records — get reflected in every run.
A fintech-focused fund cares deeply about regulatory risk and unit economics. A deep-tech fund wants heavy emphasis on IP defensibility and founder technical credibility. A consumer fund wants to understand brand signal and community dynamics. Vetrn lets you configure the agent layer to reflect your fund’s priorities — so every output is shaped by your framework, not a generic one.
The multi-agent architecture is only valuable if the output is actually usable. Here’s what we built the system to produce.
Every claim in a Vedi or Vetra output is marked as verified, unverified, or flagged as inconsistent with other known information. We never present an AI inference as a confirmed fact.
Research outputs include citations and source references. You can see where a finding came from — and follow up on it yourself if a claim is critical to your decision.
Outputs are formatted as institutional research reports — with defined sections, structured reasoning, and IC-ready formatting. Not a bullet-point summary of the pitch deck.
Every output is read through the lens of your fund’s thesis. Findings are flagged as relevant or irrelevant to your investment criteria — not presented in a vacuum.
Vedi and Vetra outputs feed directly into the Vetrn Memory layer. Every subsequent interaction — chat, meeting copilot, IC memo — has access to the full research context.
Research reports can be exported as clean PDFs or structured documents at any point. IC-formatted sections are ready to drop into your investment memo as soon as the run completes.
Book a demo and we’ll trigger a live Vedi + Vetra run so you can see exactly what multiple agents produce — and how it compares to your current research process.