Invitation-only investor diligence

Diligence should be executable.

Connect your own AI harness to inspect approved LongArena evidence. Public visitors see a bounded overview; exact work-email accounts receive MCP access only after manual review.

Request access by work emailView the Top 30 index

Global + China VC AI-Affinity Top 30

One unified public-evidence index. It does not rank investment returns, founder friendliness, fund quality, or prestige.
Version 2026.09 · audited 2026-09-19

Evidence collection, endpoint probes, and scoring for this index are performed entirely by autonomous AI agents. No per-firm human scoring is involved. Read the full evaluation methodology →

4

Sequoia Capital

Whitelist
United States@sequoiacap.comNo public endpoint found

Deep coverage of long-running agents, context engineering, and portfolio MCP deployment patterns.

Official evidence
76/100
8

MiraclePlus

China@miracleplus.comEndpoint check inconclusive

The strongest machine-readable publishing posture in the China cohort, paired with hands-on AI founder programs.

Official evidence
70/100
13

Khosla Ventures

United States@khoslaventures.comNo public endpoint found

High-conviction AI thesis with practical writing on agent workforces and AI-first company structures.

Official evidence
64/100
14

Conviction

United States@conviction.comNo public endpoint found

Purpose-built AI-native investment platform with a focused technical founder network.

Official evidence
63/100
18

Linear Capital

China@linear.vcNo public endpoint found

AI and frontier-technology specialist with a technically oriented team and portfolio.

Official evidence
59/100
19

Lightspeed Venture Partners

Global@lsvp.comNo public endpoint found

Broad agentic AI and enterprise AI coverage; public implementation evidence is less MCP-specific.

Official evidence
58/100
20

Qiming Venture Partners

China@qimingvc.comNo public endpoint found

Large AI portfolio and recurring agent-market research, with limited public harness-operation evidence.

Official evidence
57/100
21

Sinovation Ventures

China@sinovationventures.comNo public endpoint found

Long-running AI specialization and founder programs, with less public evidence of modern MCP/harness operations.

Official evidence
56/100
25

GSR Ventures China

China@gsrventureschina.comNo public endpoint found

Long AI investment history and technical portfolio depth, with little public MCP/harness operating evidence.

Official evidence
52/100
26

5Y Capital

China@5ycap.comNo public endpoint found

Strong exposure to AI founders and infrastructure, with sparse first-party evidence of internal agent operations.

Official evidence
51/100
27

Shunwei Capital

China@shunwei.comNo public endpoint found

Material AI portfolio activity, but limited reproducible public evidence of MCP or harness fluency.

Official evidence
49/100
28

Matrix Partners China

China@matrixpartners.com.cnNo public endpoint found

Broad technology portfolio and AI exposure, with limited first-party modern agent-interface evidence.

Official evidence
47/100
29

Legend Capital

China@legendcapital.com.cnNo public endpoint found

Meaningful AI investment exposure, but little public evidence of hands-on MCP/harness adoption.

Official evidence
45/100
30

01VC

China@01vc.comNo public endpoint found

Early-stage technology focus, with insufficient current first-party evidence of MCP or agent-harness operations.

Official evidence
43/100

What public visitors can see

Enough to decide whether a technical diligence session is worth your time—without exposing private customer identities, credentials, or gated financial material.

01 · POSITION

AI-native enterprise products

SalesNail, SalesTouch, and the open-source Termite Protocol—presented only at an approved, non-confidential level.

02 · PROOF

Executable diligence

Approved accounts can inspect structured facts, technical evidence, assumptions, and explicit verification gaps through MCP.

03 · CONTROL

Manual, revocable access

No public sandbox and no domain-wide login. Access is tied to one exact work email and can be revoked at any time.

First-cohort domain whitelist

These five institutions lead the current public-evidence index and their primary domains are pre-eligible for review.

IDG@idgcapital.com
a16z@a16z.com
Bessemer@bvp.com
Sequoia@sequoiacap.com
HongShan@hongshan.com

No endorsement： Name marks and domains identify whitelist eligibility only. They do not imply endorsement, investment, partnership, customer status, recommendation, or any affiliation with LongArena.

One domain is not one permission

A matching corporate domain only allows an application to enter review. LongArena approves and can revoke each exact work-email account independently.

1

Domain pre-check

The institution must be in the current first cohort or request a new public-evidence review.

2

Exact-account approval

LongArena reviews one named work email and sends a high-entropy access code only to that mailbox.

3

Real harness use

OAuth access continues only when the account demonstrates meaningful MCP usage and responsible behavior.

Access workflow

Apply from the account that needs access

The button opens your mail client. Send from your work mailbox so the reply and access code remain bound to the same identity.

WeChat may not open your work-mail app. Copy the template, switch to your corporate mailbox, and send it to demo@long-arena.com.

Only listed domains are pre-eligible. Other institutions may request a public-evidence re-review, but receive no automatic access.

After approval

Add the URL to a compatible MCP client. OAuth will ask for the exact approved email and the access code delivered to that inbox.

Supported harnesses—the same whitelist as SalesNail MCP: WorkBuddy, NamiWork, Codex, Claude, opencode, Kimi Code, EasyClaw, EasyClawCN.

https://www.long-arena.com/investormcp

Private interaction score. Continued access requires real MCP/harness behavior—not brand recognition. The private score measures standard handshake, recognized client use, tool breadth, deep diligence, resources/prompts, repeat use, and responsible behavior.
We store event categories, client name, tool/resource/prompt names, timestamps and outcomes. We do not store raw prompts or tool arguments for scoring. Meeting requests retain only the fields you explicitly submit.

How the score works

Scores use only first-party public evidence available on the audit date. Absence of public evidence is not proof of low internal adoption. A public endpoint probe checks common MCP and OAuth discovery paths; it cannot detect private systems.

25

MCP readiness

Verified public MCP/OAuth endpoints receive the highest credit; writing about MCP is scored separately.

25

Harness & agent fluency

Technical depth on agent harnesses, coding agents, evals, context engineering, and production architecture.

20

Operating adoption evidence

Public evidence that the firm or its programs use AI-native workflows, not merely invest in AI companies.

20

AI thesis & portfolio proof

Current AI thesis, specialist talent, technical programs, and relevant portfolio evidence.

10

Evidence quality

Recency, first-party attribution, reproducibility, and machine-readable publishing.

Endpoint probe scope: /mcp, /mcp/health, /.well-known/oauth-protected-resource, /.well-known/oauth-authorization-server, WordPress REST namespaces, /llms.txt, /index.md. “Not found” means no public endpoint was discovered in this bounded test as of 2026-09-19; it is not proof that no private MCP exists.

Evidence collection, endpoint probes, and scoring for this index are performed entirely by autonomous AI agents. No per-firm human scoring is involved. Read the full evaluation methodology →