Paused · wiregent.com
Wiregent Diligence
An agentic due-diligence platform that matches each founder and pitch to the VC whose own ruleset it fits. The desk is up at wiregent.com; the build paused in August 2026.
Wiregent — run the diligence you cannot run yourself.
4 profilesready-made weight profiles
Allocators and scout teams sitting on more applications than they can read. Founders with a working private build and no logo wall to show for it. Wiregent is the third party in the middle: it takes custody of the work, measures it, and returns a verdict both sides can forward.
Who it is for
The beachhead is early-stage software — companies whose product can be cloned, run, and scored.
The same spine is built to carry incorporation, insurance, and legal integrations once a deal clears.
The product
A founder connects a repository, an archive, or a vibe-coding platform.
Wiregent holds the source. A sandboxed runner clones it, measures structure and behaviour, and seals a certificate of what it saw. The investor receives the certificate and the verdict — never the code.
- A sealed evaluation. Neutral custody: the investor gets a ranked verdict, the founder keeps the source private.
- Eleven-category diligence. The allocator sets the mandate and the weights. The engine scores the pitch against that mandate, with four ready-made weight profiles.
- Forwardable certificates. Signed attestations anyone can check at /verify — no Wiregent account required.
- Agent-to-agent terms. Founder and investor agents can negotiate the deal, with a human in the loop above a scout’s limit.
- Funding that keeps watching. After a verdict, the desk can fund the work and watch what changes every day — so the next cheque is a decision, not a guess.
In the wild
The public site and the desk are live at wiregent.com.
The diligence core is a TypeScript engine, hosted in London and fronted by Cloudflare.
August 2026 added the first verification a founder did not write themselves — a UK company-registry lookup — and an assessment API so another product can call the same engine. The API is built and not yet deployed.
How it is built
React 19 and TanStack Router on the desk.
Express, Prisma, and PostgreSQL for the engine. A sealed sandbox runs and scores the private build. Claude and Gemini sit behind one model interface, with a deterministic fallback so the desk still works if a vendor is down.
Where it stands
The assessment API is built and committed, ready to deploy, so other products can call the same diligence core.
The build paused in August 2026 when Trestate became my focus. The Wiregent name moved on to my governed file agent.
Decisions and why
The investor gets a verdict; Wiregent holds the source.
WhyA ranked certificate a counterparty can trust should not require the founder to disclose. Neutral custody is the product.
The desk still works if a model vendor is down.
WhyA deterministic baseline is always on. Claude and Gemini add judgement on top.
One books layer for every movement.
WhySettlements, spend, and corrections all land in the same auditable record. Balanced books are a product guarantee, not a convention.
Technical highlights
- 086,080
Grounded in original research: a 6,080-company outcome study and ten market studies of what allocators require
- 01
Sandboxed evaluation of a private repository, archive, or vibe-coding project — the measurement never leaves the container
- 02
Signed certificates anyone can check at /verify — no Wiregent account required
- 03
Eleven-category diligence taxonomy with four ready-made weight profiles
- 04
An auditable books layer after a deal clears, so a follow-on is based on what actually happened
- 05
Vendor-agnostic models — a deterministic baseline, plus Gemini and Anthropic
- 06
UK Companies House evidence — the first proof in the system a founder did not write
- 07
An assessment API other products can call, built and not yet deployed
By the numbers.
- 11 categories
- diligence taxonomy
- wiregent.com
- live product