# [080] The downside: free add-ons around coding agents (local "firewalls," multi-agent desktop shells, token-compression proxies, open-source voice-agent bases). Funded companies are using free tiers to build enterprise funnels; individuals don't pay.

> Source: https://stringers.io/en/o/080.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> All four product types hit the top 3 of the PH daily chart and have thousands to hundreds of thousands of GitHub stars. Looks like a wave.

[Home](https://stringers.io/worth/) / [080]

# Not promising: free add-ons for coding agents. Local 'firewalls', multi-agent desktop shells, token-compression proxies, open-source voice agent bases. Funded companies give them away free to build enterprise funnels, and individuals don't pay.

Close Inspirational (downside) · Looks hot, but it runs on things small teams don't have (funding, enterprise sales) First seen 2026-11-01 · Last updated 2026-11-04

Status

Close (not recommended to follow: individual developers don't pay, enterprises need compliance and sales, funded companies are using free tiers to build funnels, and vendors reset everything with an update).

Why this is written up

All four product types placed top-3 on Product Hunt's daily chart and have several thousand to over a hundred thousand GitHub stars. They look like a trend. But among the PH sample, 7 companies with backing are all free, and not one has revenue evidence. The next step is all enterprise: MDM distribution, team editions, a $300/month Team tier, private deployments for BFSI. A small team following this hits the same wall: individuals don't pay, enterprises need compliance and sales, and you don't have $2M.

Confidence

Medium-high. Conditions that would overturn this (when it becomes doable for a small team): an indie interceptor, compressor, or multi-agent shell built by a team of one or two gets over $5k in revenue in 30 days and doesn't drop for two straight months; or Cursor / Claude Code enterprise editions start requiring third-party agent audit reports and a self-serve, per-seat paid tier appears.

## Evidence

Grouped by store. Each group starts with the charts that can be graphed (real curves, data dates below the chart), then the scanner's original notes and source links.

### GitHub4 evidence · 5 products with curves

[hardenrun/aif](https://github.com/hardenrun/aif)

GitHub · created 2026-08-01 · 1,866 stars · 7 days +61 · [see product profile](https://stringers.io/worth/p/agent-integrity-foundation-aif.html)

On device Firewall for Coding and Personal Agents

Stars

Data 2026-10-09 → 2026-11-08 (31 days, 31 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

[generalaction/emdash](https://github.com/generalaction/emdash)

GitHub · created 2025-08-28 · 6,214 stars · 7 days +83 · [see product profile](https://stringers.io/worth/p/emdash.html)

Emdash is the Open-Source Agentic Development Environment (🧡 YC W26). Run multiple coding agents in parallel. Use any provider.

Stars

Data 2026-09-01 → 2026-11-08 (69 days, 13 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

[JuliusBrussee/caveman](https://github.com/JuliusBrussee/caveman)

GitHub · created 2026-04-04 · 118,147 stars · 7 days +1,637

🪨 why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.

Stars

Data 2026-07-30 → 2026-11-08 (102 days, 18 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

[DietrichGebert/ponytail](https://github.com/DietrichGebert/ponytail)

GitHub · created 2026-06-12 · 186,299 stars · 7 days +4,822

Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.

Stars

Data 2026-07-30 → 2026-11-08 (102 days, 20 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

[dograh-hq/dograh](https://github.com/dograh-hq/dograh)

GitHub · created 2025-09-09 · 6,038 stars · 7 days +41 · [see product profile](https://stringers.io/worth/p/dograh.html)

Open source voice AI platform. Self-hosted alternative to Vapi and Retell. On Prem, BYOK across Speech to Speech or LLM/STT/TTS, with a visual workflow builder

Stars

Data 2026-10-28 → 2026-11-08 (12 days, 12 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

- GitHub (2026-11-02): hardenrun/aif has 1,805 stars and 392 forks, growing 0 for 6 straight days since 10-27 (repo marked "cooling"). The repo contains only README / LICENSE, 8 commits, 25 KB, and no source code. Of the latest 100 forks, 97 came from new accounts registered on the same day 2026-08 (archive inference points to an account farm). Behind it is VizopsAI, with $2M pre-seed (NewsBytes web report, single source).

- GitHub (2026-11-02): generalaction/emdash at 6,141 stars, adding 8–10 stars a day in a straight line since its PH launch (2026-05-20,405 votes). YC W26, desktop app free, team plan 2.0 is waitlist-only. The site claims "1M+ downloads," but broken down, that's update checks: of v1.2.8's 170,489 downloads, only 2.7K were installers (from the GitHub Releases API on the web, see archive 11-01).

- GitHub (2026-11-02): JuliusBrussee/caveman 116,663 stars (+8,380 in 30 days), DietrichGebert/ponytail 182,027 stars. The "save tokens" slot is taken by two free repos. Caveman 10-13 became a YC F26 company and launched Cloud Team at $300/month. There's no indie library entry for caveman.so and no revenue evidence in two weeks. HN 904 points (id 47647455), but the skeptic thread with 89 points (id 47954745) measured real task output tokens dropping only 8.5–9%.

- GitHub (2026-11-02): dograh-hq/dograh has 6,005 stars and 1,537 forks (fork/star ratio 25.6%, users are "taking it to self-host"). The Docker image has about 87k pulls (web Docker Hub). PH 583 votes, daily chart #1 (2026-08-12). The cloud version is 1¢/minute. No dograh.com for indie developers, and zero public revenue evidence.

### Product Hunt1 pieces of evidence

- PH (2026-09-09): Harden 369 votes (379 today), daily chart #2, 98 comments. Over the same period, Show HN got only 9 points (id 49472151). The personal edition is "free forever." The enterprise edition goes through MDM / Jamf distribution plus contact sales. Not a single revenue or user figure can be found.

### Indie product revenue (TrustMRR)1 evidence · 2 products with curves

[Headroom](https://trustmrr.com/startup/headroom) extraheadroom.com

TrustMRR · 30-day revenue $3,068 (+6%) · MRR $2,233 · 251 subscriptions

Monthly revenue (verified by payment platform)

2026-04 → 2026-11 (8 pts). Light bars = current month, not yet complete

Daily revenue (UTC day)

Data 2026-09-29 → 2026-11-06 (39 days, 39 observations). Single-day swings are big, so look at the trend in the monthly breakdown above.

[Token Cut](https://trustmrr.com/startup/token-cut) tokencut.lol

TrustMRR · 30-day revenue $0 (+0%)

Daily revenue (UTC day)

Data 2026-09-29 → 2026-11-06 (39 days, 39 observations). Single-day swings are big, so look at the trend in the monthly breakdown above.

- indie (30-day revenue, as of 11-01): the small teams in this space with visible revenue are tiny. Headroom (menu bar app that cuts Claude Code spend) $2,962, Token Cut (tokencut.lol) $0.

## Already doing it

Harden (VizopsAI), [Emdash](https://stringers.io/worth/p/emdash.html) (YC W26), Conductor, Codex App, Superset; Caveman (YC F26), Ponytail; [Dograh](https://stringers.io/worth/p/dograh.html), LiveKit, Pipecat. Claude Code, Codex and Cursor are also adding permissions, sandboxes and parallel sessions themselves. [012] (security scanning for Skills / MCP, free and open source from big vendors) is the runtime version of the same conclusion; [004] (usage tracking) is in the same coding-agent periphery. All of them crowd into spots vendors will fill themselves.

In our product dossiers: [Harden (AIF / Agentic Integrity Foundation)](https://stringers.io/worth/p/agent-integrity-foundation-aif.html), [Emdash](https://stringers.io/worth/p/emdash.html), [Dograh](https://stringers.io/worth/p/dograh.html)

## The hardest step

Paid. Free and open source is the default option; enterprise sales cycles are the only moat, and one vendor update can make the third-party layer redundant.

## Who it's for

Not a good fit for a small team alone. Open spots: the "adversarial (prompt injection) detection" that Harden has carved out on its trust page needs security research background; bundle by industry on [Dograh](https://stringers.io/worth/p/dograh.html) / Pipecat (hospital booking, restaurant phone answering) and sell directly to SMBs; a lightweight compressor for individuals and small teams that takes a cut of the savings (inferred).

## Should every product builder know this?

Yes. PH votes and GitHub stars are the easiest demand signals to misread in this category, and Harden also showed stars and forks can be faked.

## Change log

2026-11-04

PH (PT 11-02) day chart #1: beacon.sh (406 votes) = GitHub Asymptote-Labs/agent-beacon (open-source self-improving memory layer for coding agents, 3,227 stars); the two match, but it's open source and free. t3code has 43,970 stars, +7 in 13,036 days, and is still accelerating. Free periphery keeps squeezing the paid space.

2026-11-03

pingdotgg/t3code has 41,288 stars, +7 in 10,657 days (+3,841 today). The strongest free coding-agent desktop shell is still accelerating, further squeezing the paid competitor space. Also Balanov/uncut (7MB static binary coding-agent harness, 2 days, 570 stars).

2026-11-02

Backfill check: hardenrun/aif stars unchanged at 1,805 (from 10-27 onward growing 0); emdash 6,131 → 6,141; caveman 116,510 → 116,663; dograh 5,997 → 6,005; ponytail reported ~181k, today 182,027. Still can't find any indie revenue for Harden, [Emdash](https://stringers.io/worth/p/emdash.html), Caveman or [Dograh](https://stringers.io/worth/p/dograh.html). Conclusion unchanged.

2026-11-01

First recorded (backfill: from Product Hunt Roundup v1 4.2 item #5, ph-dossiers harden / emdash / caveman / dograh, and Stringers section 5 "two paths: solo or funded company", data as of 2026-11-01).

Original: Scanner opportunity library #080 (updated daily, 06:00); this page was generated on the 2026-11-08 edition.