1,229 weekly PyPI installs. 0 paid signups.

that ratio sat for 3 weeks. the natural read was distribution, so I shipped a Smithery listing, OpenAPI spec, agent starter repo, public demo. install count kept climbing. conversion stayed at zero.

ratio held. the bottleneck was not reach.

shipped v0.3.2 today to find out if the gap was the free tier itself. before this release, free tier returned the same data shape as Pro for most tools, just with a 24h cache lag. a curious dev could install, get useful answers, and never hit a wall. no wall, no reason to pay.

v0.3.2 narrows the free tier:

• 3 tools available (top_yield_farms, hl_vault_leaderboard, macro_tape)

• 5 results per query

• 60 req/hour

• no signup

Pro at $19/mo opens the other 10 tools, 100 results per query, real-time refresh, 90 days of history on equity feeds. Pro Plus at $49/mo adds webhooks.

the gate matters because of which tools moved. confluence_today is now Pro-only. it stacks 2+ signals on the same asset across SEC form 4 filings, ETF flow, polymarket whale positions, and on-chain wallet activity. it is the only tool in the lineup that does cross-source alignment. on free tier it now returns a stub naming the tool and the /pro link, instead of yesterday’s data.

three radars also moved to Pro:

• earnings_drift_radar: post-earnings drift candidates with IV-crush filter

• token_unlock_radar: forward unlock schedule + dollar-weighted size

• fed_comm_radar: Fed speaker windows with historical realized-vol bands

these shipped in v0.3.0 but were on the free tier by accident. v0.3.2 closes that.

the in-tool upgrade nudge is the other half. every free-tier response now includes an inline note pointing at /pro, and every Pro-only call on a free key returns a stub with the tool name and the link. first time a Cursor or Cline or Windsurf user running the MCP will see the Pro path surfaced inside their own agent’s response, with the data they wanted instead of around it.

it does not interrupt the workflow. it does not nag on every call. one note in the response, one stub on gated calls.

the funnel test now has signal. v0.3.2 ships telemetry on per-tool calls and has_key. we can measure: of N free-tier installs, how many call a Pro-only tool? of those, how many convert in 72 hours? the rate is the answer.

I do not know what the rate will be. the prior data (1,229 / 0) suggested converted-conditional-on-wall-hit was simply untested, not zero. it could still be zero. if it is, the funnel is broken at a deeper layer than gating, and the next move is a 14-day trial, a smaller annual price, or pulling Pro out and going B2B.

if it is 1%, the math works. 1,229 installs a week, 1% conversion at $19 average revenue per user is about $230/wk. small but real. doubles if reach doubles.

if it is 5%, distribution is the new bottleneck and the work goes back into that channel.

here is what I did not do this release. I did not bundle the 3 new radars as separate launches even though they are launchable. I did not inflate the metrics (“trending on PyPI”, “growing”, “join the buzz”). there is no buzz. there are 1,229 installs and zero customers and a hypothesis that needs a test.

72 hours from now the telemetry will say whether the gate converts. I will post the number.

pip install falsifylab-alpha-mcp==0.3.2

falsifylab.com/pro?ref=tool-nudge

#OnchainAlpha #DeFiYield #StablecoinYield

— research and educational content. not investment, legal, or tax advice. do your own research. positions and views may change without notice.

Originally published on FalsifyLab Substack.