GoatCounter + GitHub · open-agent-ai-security

Community Traffic — Overview

Reach combines our hero websites (GoatCounter) and GitHub repo traffic, per project, for 2026-06-17 → 2026-09-29. Real traffic only — automated email-campaign hits are tracked separately (see Marketo campaign tracking), and both sources undercount humans (blockers, no-JS), so true reach is if anything higher.

14,094Total reach

4Projects tracked

105Days in window

PraxenTop project

Reach by project

Pages-site views and repo views, combined. Community = the shared root / cross-project pages (no repo of its own).

Community1578 · 11%

Praxen6979 · 50%

Observra4016 · 28%

Promptfall1217 · 9%

Raffkin304 · 2%

Daily views — site

Community sites · real views per day

━ 7-day average · ▮ daily views · ▲ key date.

Total reach — all-time

Community sites (views) + GitHub repos · cumulative reach

━ cumulative total · ▲ milestone.

Top referrers

Pages-site + repo referrers, combined.

Direct (no referrer link)5474 · 39%

Internal site nav1992 · 14%

LinkedIn1128 · 8%

Search621 · 4%

MS Teams334 · 2%

Exabeam.com308 · 2%

Other / press211 · 1%

GitHub.com111 · 1%

Facebook27 · 0%

X/Twitter12 · 0%

Top pages — site

Praxen `/praxen`	3755
Observra `/observra`	1696
Community `/`	1066
Promptfall `/promptfall`	1032
Praxen `/praxen/guide/quickstart.html`	295
Praxen `/praxen/guide/index.html`	232
Observra `/observra/guide/index.html`	167
Praxen `/praxen/news.html`	158
Praxen `/praxen/guide/installation.html`	133
Observra `/observra/demo`	125
Praxen `/praxen/guide/abv.html`	123
Community `/blog`	112

Top locations — site

California, US2222

Virginia, US1005

Washington, US732

Iowa, US576

India562

United Kingdom552

Texas, US459

Singapore454

Colorado, US442

Australia440

Netherlands411

Germany353

Japan334

Ireland300

Canada222

GitHub stars

A GitHub star is a public bookmark — a developer flagging a project they find useful or want to follow. It's the open-source equivalent of a “like,” and the running total is the standard rough proxy for a project's mindshare and momentum. Live counts per project below, with the combined trend over time.

★ 61Praxen · open-agent-ai-security/praxen

★ 23Observra · open-agent-ai-security/observra

★ 7Promptfall · open-agent-ai-security/promptfall

★ 6Raffkin · open-agent-ai-security/raffkin

Cumulative stargazers from the GitHub API. A launch is a spike; the slope over the following weeks is the real adoption signal.

PyPI downloads · Observra

Observra only. Installs of the observra Python package via pip, from the public pypistats.org dataset — Praxen ships as a Claude Code / Codex plugin, not a pip package, so it has no PyPI footprint. Counts are all-inclusive (they include CI and mirror pulls), so the smoothed line, not any single spiky day, is the signal.

1,209Observra · last 30 days

207last 7 days

7,424tracked total

Observra observra · downloads per day

━ 7-day average · ▮ daily downloads · ▲ version ship.

Total installs — Observra

Observra · cumulative pip downloads

━ cumulative total · ▲ milestone.

Praxen plugin installs

Praxen only. Praxen installs as an agent plugin (Claude Code / Codex), and those marketplaces expose no public download counter — so we use GitHub repo clones as a proxy for installs (137 unique installs in the trailing 14 days, GitHub-deduplicated). It is a proxy: the count also includes update re-pulls and CI, so — like the PyPI numbers — read the trend, not an exact headcount. (Observra ships on PyPI, hence its different install signal.)

514Praxen · installs, last 30 days

145installs, last 7 days

4,308tracked total installs

Praxen open-agent-ai-security/praxen · installs per day (clone proxy)

━ 7-day average · ▮ daily installs · ▲ release / launch.

Total installs — Praxen

Praxen · cumulative plugin installs (clone proxy)

━ cumulative total · ▲ milestone.

Marketo campaign tracking

Marketing email blasts are auto-detected here whenever tokenized email traffic (Marketo mkt_tok links) spikes past 50 hits/hour. Almost all of that volume is corporate email-security infrastructure (Microsoft Safe Links, Mimecast, Proofpoint, …) auto-fetching every link — not people — so we hold it out of the reach figures above and log each send's real human click-through below. New campaigns append here as they land.

Detected campaign	Tracked	Human	Scanners	CTR
📧 2026-07-01 · Marketo blast	11,087	121	10,966	1.1%

Human click-throughs (est.)121 · 1.1%

Automated scanner fetches10966 · 98.9%

A ~1.1% human-signal rate — recipients showing a mobile viewport or a time-separated revisit — is a healthy click-through for a B2B email send; the rest is unavoidable scanner traffic every blast generates. How each pageview is attributed (Marketo token → mobile / revisit / burst-hour screen) is documented at the top of stats/generate.py.

Search-driven traffic

Views whose referrer is a search engine (Google, Bing, DuckDuckGo, Brave, …). Like all referrer data this undercounts — many search visits arrive with no referrer — so read the trend, not the absolute level. The marker is when the SEO/GEO work went live; organic search indexes and compounds over weeks, so lift shows up after it, not immediately.

Community sites · views referred by a search engine, per day

━ 7-day average · ▮ daily search views · ▲ SEO/GEO work live.

Generated 2026-09-30 by stats/generate.py. Sources: GoatCounter export (Pages sites, community account) · GitHub repo traffic API (repo views/clones, 14-day window) · stars via GitHub API. Reach (14,094) = 10,075 real site views (2026-06-17→2026-09-29) + 4,019 repo views (accumulated daily, kept past GitHub's 14-day purge); a further 10,966 automated email-campaign scanner pageviews were tracked separately (see Marketo campaign tracking). Method documented in stats/generate.py.