I Tracked AI Tools for Crypto Builders for 2 Weeks: What Actually Held Up

Two weeks ago I set out to answer a boring but expensive question: which "AI-powered" research tools actually change the outcome of crypto due diligence, and which ones just make you feel productive while you scroll.

I ran a rotating stack against real DD tasks — vetting a new founder before an OTC deal, checking a "audited" protocol's claimed team, sanity-checking a airdrop farm before committing a wallet. Most tools I tried were LLM wrappers around search results you could've gotten from Google in the same time. One tool wasn't: Maigret, an open-source username enumeration tool with 26,000+ GitHub stars that checks a single handle against 3,000+ sites in under a minute.

It's not flashy. It doesn't have a chat interface. It's exactly the kind of unsexy tool that quietly does the job the AI wrappers were pretending to do.

What I actually tested, and why most of it flopped

The pattern with most "AI research assistant" tools for crypto is: paste a wallet address or a Twitter handle, get a paragraph of LLM-generated summary that sounds confident and contains almost nothing you couldn't have found in 90 seconds of manual searching. They're good at narrative, bad at coverage. For DD, coverage is the whole point — you're trying to find the one account, the one old post, the one abandoned GitHub repo that tells you this founder has burned two projects before under a different name.

Maigret does the opposite. No narrative, no summary, no LLM in the loop at all — it's straight enumeration. Feed it a username, it checks that string against thousands of site templates (GitHub, Telegram, Reddit, niche forums, old blogging platforms, crypto-specific communities) and reports back every place an account with that handle actually resolves. It's blunt instrument work, and that's exactly why it's reliable — there's no model hallucinating a plausible-sounding but wrong result.

Worked example: vetting a founder handle

Here's the exact workflow I ran on a founder pitching a "stealth" DeFi protocol, using a placeholder handle since the real one isn't mine to publish.

1. Install it (one line, no API key required):

pip install maigret

2. Run the scan against the handle they gave me on their pitch deck:

maigret cryptofounder_handle --timeout 10 --html report.html

3. Let it run — this took about 45 seconds against the default site list (~3,000 sites; you can widen it further with --top-sites 0 to hit every site in the database, which takes a few minutes longer).
4. Open report.html. It lays out every confirmed match with a link, grouped by category (social, crypto, coding, forums).

The result: three accounts I'd never have found manually — an old Reddit account from 2019 posting in a scam-adjacent subreddit under the same handle, a GitHub with commit history to a project that rugged in 2021, and a Telegram presence in a group associated with pump-and-dump calls. None of that showed up in the first four pages of Google. All three took one 45-second command to surface.

That's the whole pitch. It's not AI in the "reasoning" sense — it's automation doing exhaustive, boring lookup work that a human would take hours to replicate by hand, and would probably give up on after the first ten sites.

Maigret vs. the alternatives

| Approach | Time for full pass | Coverage | Best for | |---|---|---|---| | Manual search (Google, X search, LinkedIn) | 1-3 hours | Whatever ranks in top results | Quick gut checks, low-stakes follow decisions | | LLM research wrapper (ChatGPT/Grok w/ browsing) | 5-10 min | Narrow, biased toward indexed/recent content | Summarizing what's already public and easy to find | | Maigret | Under 1 min | 3,000+ site templates, including obscure/old platforms | Pre-deal DD, background checks on unfamiliar handles | | Paid OSINT platforms (Pipl, Social Catfish, etc.) | Minutes | Similar or narrower, plus paywalls | When you need identity resolution beyond usernames (real name, address) |

The honest tradeoff: Maigret only works if you have a consistent username to search. If your target uses a different handle on every platform, you get nothing — it's not doing semantic matching, just exact-string lookups (with some variation tolerance). For that scenario you're back to manual work or a paid identity-resolution service, which is a different category of tool entirely and overkill for routine DD.

It's also worth saying plainly: this isn't a replacement for reading a whitepaper, checking a multisig, or calling references. It's one input into DD, not the whole process. Treat a clean Maigret scan as "no red flags found," not as a clearance.

Common mistakes

• Running it once and calling it done. Handles change. Rerun the scan periodically for active relationships (an ongoing OTC counterparty, a team you're vesting into), not just at intake.
• Ignoring negative results. A handle that resolves to almost nothing across 3,000 sites is itself a signal — either it's a brand-new persona, or the real person is careful about reuse. Both are worth noting, not shrugging off.
• Skipping the --timeout flag on slow connections. Default timeouts can cause false negatives on legitimately-existing accounts if your connection is slow — bump it to 10-15 seconds rather than trusting a fast scan's negatives at face value.
• Treating a single match as confirmation of identity. Usernames collide. A match is a lead to manually verify (check post history, profile photo, join date), not a fact to act on directly.

What's rising and what's fading

The trend over these two weeks was clear: LLM-wrapper research tools are becoming background noise — there are too many of them, they mostly do the same shallow web search, and crypto Twitter is (correctly) getting more skeptical of anything that outputs a confident paragraph without showing its receipts. What's gaining traction instead is narrow, deterministic tooling that does one unglamorous job exhaustively — enumeration, on-chain tracing, contract diffing. My bet for the next cycle: DD workflows that chain a few of these narrow tools together (username enum → wallet clustering → contract verification) will outperform any single "all-in-one AI research" product, because each step is auditable instead of a black box.

FAQ

Is Maigret legal to use for background checks?

Yes — it only checks for public account existence across sites using a username, the same information you could find by manually visiting each site. It doesn't scrape private data or bypass access controls. Standard OSINT ethics still apply: use it for legitimate DD, not harassment or stalking.

Does Maigret work if someone uses different usernames on different platforms?

No. It matches on the exact (or near-exact) username string, so it misses any account where the person used a different handle. For that case, cross-reference other signals — profile photos, bios, linked accounts — using manual search.

How is this different from just Googling a username?

Google indexes a fraction of the sites Maigret checks directly, and it ranks by SEO relevance, not existence. Maigret hits the actual site with a direct request for that specific username, so it surfaces old, low-traffic, or deliberately unindexed accounts that never show up in search results.

---

→ Ask the index what to build your web3 ai stack

→ Free credits for these tools

Written by McKlaud AI. Want to know which AI tools actually fit your business? Get a free AI audit.