callstackincubator/simlock · callstackincubator

One agent, one simulator: Simlock gained 368 stars in a day, and it wasn't the idea that did it

What it is An open-source command-line tool plus a resident process that rents simulators and real devices to multiple coding agents working at the same time, one agent per device, reclaimed after use, and can also link several computers into a single pool.

Why it took off Parallel multi-agent coding has become the norm, and mobile was the first to hit the problem of simulators being fought over. The Callstack team, its CEO, and the hot client T3 Code all promoted it on the same day, and the stars clustered on launch day.

CHow you could follow Don't build a second Simlock. Copy its topic selection and launch sequence: find a new friction point where "multiple agents fight over the same local resource" (browser login state, ports, test databases, etc.), build it fully in private first, and before launch get a hot agent client to integrate it first, with the team reposting the same day.

Grade C: the traction mainly came from Callstack's distribution and T3 Code's launch integration, which followers can't get. Projects with the same concept but no distribution only reached 1 points and 127 stars on Hacker News, so copying it directly isn't worth it. What you can learn is the topic-selection method and the launch playbook. Being talked about in small circles

Data (updated daily) · Repo created 2026-07-17 · 619 stars · 7 days +96: GitHub · stars
Data 2026-11-01 → 2026-11-08 (8 days, 8 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

Szymon Chmal of Callstack released Simlock in Oct 30. It solves a narrow problem: when several coding agents write mobile code at the same time, stop them from fighting over the same iPhone simulator. On launch day it got 368 stars, and a week later 610. Someone had open-sourced the same idea back in September, and it only got 1 points on Hacker News. The gap between the two wasn't the code; it was the people standing by to help on launch day. Below I explain what it is, where the stars came from, and which parts to copy if you want a project that gets stars the same way.

The night before launch, someone already said "it works"

On Oct 29 22:16 (UTC), Julius (@jullerino) of T3 Code posted on X: "You can now use Simlock in T3 Code," and added that it's a "great alternative" to EAS Simulator, Expo's paid simulator cloud, "if you happen to have some spare [machines]." Simlock hadn't officially launched yet. T3 Code is an agent client built by Theo's team; the same week it had 48k stars on GitHub, gaining 16k over 7 days, right when the most people were watching.

12 hours later, on Oct 30 10:41, the author posted the official launch tweet: "Introducing Simlock: a free, open-source tool that turns your machine into a simulator cloud that manages itself, so every coding agent gets its own iOS or Android device, simulator or real phone." The official Callstack account replied below with "let's go, self-host everything," and CEO Mike Grabowski (@grabbou, whose bio lists being an OpenAI Codex ambassador) reposted it that day, and posted again on Nov 3. Comments included "great launch as always." This group has a fixed launch routine, and a set of viewers who show up every time.

What it actually is

An npm package plus a resident process. Once installed, when an agent needs to test an app it first "rents" a device:

``sh npm i -g simlock simlock lease --platform ios --device "iPhone 16" ``

The device it gets is already booted and checked, and no other agent can touch it during the lease. If none is free, one is created on the spot; when the machine is full, requests queue, and unused devices are reclaimed automatically. iOS simulators, Android emulators, and registered real devices all go through the same lease system. If one Mac isn't enough, a gateway can link several Macs and Linux machines into one pool with a single external address. Access options include a command line, MCP, an HTTP API, a Node client, and a web console. MIT license, and the README says it "will always remain free."

The problem it solves can be stated in one sentence: when several agents run at once, simctl and adb see the same set of devices, so one is installing an app while another wipes the simulator. A more blunt comment under the launch tweet put it this way: "Two agents on one sim is how a test taps a button the other agent just deleted."

How the stars came

Date	Stars	Source
10-30 launch day	368	The author's own account (X / LinkedIn, 11-03)
11-01	523	GitHub, logged daily
11-03	561	Same as above
11-06	605	Same as above; in the second week after launch it kept gaining 10~20 per day

npm daily downloads show the rhythm more clearly: before launch it was single digits per day; on Oct 29 36 it jumped to 30 on Oct 474 it hit 5 and then stayed at two or three hundred a day; the day v0.6.0 came out in Nov it fell back to 476 (npm downloads API). More than half the stars landed on the launch day, which suggests it wasn't found slowly through search; a coordinated launch push got it going. It never appeared on GitHub trending.

Four things stacked on the same day:

A partner integrated it first. T3 Code announced it worked the night before. Callstack's people replied under that tweet: "Thanks for being the earliest adopter! … give it star on GitHub."
The team launched together. The author posted a video, the CEO and the official account relayed it, and LinkedIn posted it too. On day 4, the author posted another round with "500+ stars, 368 of them on launch day; 1,000 npm downloads in the first three days; 90+ new issues and 70+ merged pull requests."
The author already had a reputation. Szymon Chmal built React Native tools like Rozenite, Harness, and Voltra, is a Microsoft MVP, and gave a talk at Chain React titled "Giving agents hands and eyes in React Native." When he posts, people in the RN community already pay attention.
The README asks for stars directly: "If you think it's cool, please star it 🌟."

Within what I could find, there was no Hacker News thread, no Reddit discussion, and no media coverage. Its spread stayed almost entirely within the RN and X agent-tools circles.

The same idea, built by others earlier

Project	Timeline	Result	How it differs from Simlock
BariBariGood/manzanas	2026-09-01 Show HN	1 points, 0 comments	Nearly the same design: a resident Mac process, leases, a warm pool, multiple Macs, MCP, iOS only
entropyconquers/simfleet	2026-09-25	127 stars	Only macOS + React Native, with Metro ports and build caching
AgentCloud	2026-08-27 Show HN	1 ratings	Hosted iOS simulator cloud
EAS Simulator(Expo)	2026-09-14 preview	Scale not disclosed	A paid hosted cloud with replay. Simlock targets the opposite: self-hosted, free, using the Mac you already have sitting idle

Several others are complementary rather than competitors: MobAI-App/simslim (3,499 stars, lets one Mac run several simulators; someone under the launch tweet already asked whether Simlock would adopt its tricks), mobile-next/mobile-mcp (10k+ stars, lets agents "operate" phones), and Callstack's own agent-device (5,277 stars, also an operation layer, a natural pairing with Simlock, one handling allocation and the other handling operation).

manzanas came out two months before Simlock with nearly the same design, and nobody noticed. So Simlock's broader scope (iOS + Android + real devices, clustering, multiple access methods) is real, but that decides whether people are willing to star and open issues, not whether anyone shows up at all. The people came from Callstack. A week after launch, no new similar project has emerged yet.

Still unclear

How many of the 610 stars belong to people actually using it can't be determined. The npm downloads include CI and mirror traffic.
How many of the "90+ issues and 70+ PRs" came from outside contributors hasn't been verified. The commit history is almost entirely the author alone.
The T3 Code integration is still behind a Labs toggle, and how many users it actually brings is unclear.
The like and repost counts on those tweets haven't been verified.

If it were you

C. You'll learn a launch playbook from this, but copying it to build an "agent device rental" simulator will probably not get you stars. The gap is already taken by a company that has distribution. The two without distribution, one has 127 stars, the other 1 points.

Who it fits: people who already have a distribution channel in the agent tooling crowd. Say you have a few thousand followers on X who build with agents, or you can get a popular agent client (T3 Code, Cline, OpenCode and the like) to integrate it the day before you launch. Also people already building developer tools for an ecosystem who want to add an "for agents" piece.

Who it doesn't fit: solo devs who want an iOS/Android simulator scheduler as their first open-source project and expect the README and a Show HN to bring traffic. That's exactly what manzanas did, and it got 1 points.

How to do it: don't build a second Simlock, copy how it picked its topic. Its framing is "when multiple agents run in parallel, they fight over the same local resource." Simulators are just the first place mobile devs hit it. The same friction shows up elsewhere (below is inference, I didn't check each one for existing projects): browser profiles and logged-in sessions, local ports and dev servers, test databases, Docker resources, GPUs. Pick one you keep getting burned by yourself, make it "one per agent, reclaimed when done," give it a name that's clear at a glance, and state the pain in one sentence. If you want to sit next to Simlock, build the parts it hasn't: turn SimSlim's trick of running several simulators on one Mac into a plugin for it, or make a framework-specific version like simfleet did for Flutter. Projects like this will get stars an order of magnitude lower, but they can ride Simlock's traffic.

How to get users: follow its launch order. Privately get the docs, the console, and every integration path ready first (it created the repo in July, put it on npm in August, and only went public at the end of October). Before launch, get a popular client to integrate it, and have them post the night before. On launch day, the author posts a tweet with a demo video, and people who know them retweet the same day. Pick a paid cloud product as a foil and pitch "self-hosted, free." On day 4, post again with the star and download counts. Ask for stars directly in the README.

When before what?: the wave of infrastructure for "parallel multi-agent" work is happening right now. My guess is that by the first half of 2027, each agent client will build isolation and leasing into itself. At that point a standalone scheduling tool won't be needed (inference).

Ways this could be wrong: ① If two or three "agent resource rental" projects with no big-company backing and a few hundred stars each appear within a month, the category is growing on the idea alone, and the grade should go up. ② If Simlock is still adding double-digit stars a day in December and most PRs come from outside, it's a real need in use, and building plugins and add-ons for it is worth more. ③ On the other hand, if T3 Code, Codex, or Claude Code ships device isolation built in, or if Simlock is adding only single digits a day a month from now, this category is finished.

Sources: GitHub project page and daily star counts (11-01 to 11-06), npm registry and download API, the author's and T3 Code's posts on X (10-29, 10-30, 11-03), the author's LinkedIn post (11-03), the Simlock README and docs site, the Expo EAS Simulator page, and the Show HN posts for manzanas and AgentCloud on Hacker News.

GitHub repo ↗

Research notes: competitor table, acquisition evidence, full research report

Competitors

Name	Where	Launched	Scale	How it differs from this one
EAS Simulator(Expo)	Big-company features	2026-09-14 preview	Expo ecosystem, not public	Paid hosted cloud + session replay. Simlock is self-hosted, free, and runs on your own Mac.
BariBariGood/manzanas	Open source	2026-09-01 (Show HN 1 points)	Very small	Nearly the same idea (lease, warm pool, multiple Macs, MCP), but iOS only, and no distribution
entropyconquers/simfleet	Open source	2026-09-25	127 stars	Tied to macOS + React Native, with Metro ports and build caching. Simlock isn't tied to a framework and also supports Linux, real devices, and a gateway.
AgentCloud	Web	2026-08-27 (Show HN 1 points)	Very small	Hosted MCP iOS simulator cloud
MobAI-App/simslim	Open source	2026-07-20	3,499 stars	Complementary: lets one Mac run more simulators, doesn't handle allocation
mobile-next/mobile-mcp	Open source	2025-03	10,317 stars	Lets agents operate devices, doesn't handle isolation across multiple agents
callstack/agent-device	Open source	2026-01	5,277 stars	Same company's device-operation layer. Simlock handles allocation, and the two work as a pair.
avito-tech/Emcee, pxctest, GADS, and others	Open source	2016–2022	200–1,000 stars	CI test sharding and device farms, not interactive device leasing for agents
BrowserStack / AWS Device Farm / Firebase Test Lab	Other	Already around	Big	Paid cloud real devices, aimed at QA

Where users come from

Verified launch-partner integration (T3 Code): @jullerino posted "You can now use Simlock in T3 Code" on 2026-10-29 22:16 UTC, about 12 hours before the official launch. t3code was at 48k stars in the same period, +16.5k over 7 days.
Verified X founder/company network: author 10-30 posted a video; CEO @grabbou (OpenAI Codex ambassador) promoted it 10-30 and 11-03 times; @callstackio's official account amplified it
Verified author's personal brand: Szymon Chmal is the author of Rozenite, Harness, and Voltra, a Microsoft MVP, and a Chain React speaker. Comments include "great launch as always."
Verified LinkedIn: both the author and the CEO posted, using the numbers for a second round of spread.
Verified README asks for stars: "If you think it's cool, please star it." Replies also ask people to click the star.
Inference HN/Reddit: neither the hackernews library nor a Reddit search turned up anything.

Why callstackincubator/simlock took off

In one line: Simlock is an open-source CLI and daemon from Callstack, the React Native consultancy and open-source company. It "leases" iOS simulators, Android emulators, and real devices to coding agents running in parallel: each agent gets one, and it's reclaimed when done. When one Mac isn't enough, a gateway can pool several Macs or Linux machines. Takeaway: it hit the pain point that just emerged, where multiple coding agents writing mobile code in parallel fight over simulators. The initial traction came from Callstack's own X network plus the T3 Code launch integration, a coordinated release (368 stars on launch day). The product differences (self-hosted, iOS + Android + real devices, a gateway for fleets) are real, but someone already built the same idea first, so the stars came mainly from the publisher's distribution.

1. Basic data

Repo created on 2026-07-17 (developed privately first), npm 0.1.0 published on 2026-08-21 (registry.npmjs.org/simlock), official public launch 2026-10-30 (Friday): author Szymon Chmal posted "Introducing Simlock" at 10:41 UTC that day (x.com/ChmalSzymon/status/2116118360270495865).
Numbers the author shared (LinkedIn / X, 2026-11-03): "500+ GitHub stars, 368 of them on launch day; 1,000 npm downloads in the first three days; 90+ new issues and 70+ merged pull requests" (linkedin.com/posts/szymon-chmal_four-days-ago-we-launched-simlock…).
Our GitHub library repo_day: 11-01 523 → 11-02 543 → 11-03 561 → 11-04 570 → 11-05 588 → 11-06 605. In the 2 week after launch it kept gaining +10 to +20 a day, peaking at +20 (11-02). GitHub API live: 11-06 610 stars, 30 forks, 68 open issues. It never made trending: trending only counts repos 60 days after creation, and this one was created early and made public late, so that rule didn't apply to it.
npm daily downloads (api.npmjs.org): 10-29 36, 10-30 474, 10-31 246, 11-01 280, 11-02 199, 11-03 153, 11-04 323, 11-05 476 (v0.6.0 released that day). Before launch it was single digits per day.
Versions: v0.4.0 (10-30), v0.5.0 (11-01), v0.6.0 (11-05), three releases in one week. Nearly all commits are from Szymon Chmal alone.
GitHub hides the stargazer detail endpoint (REST 404, GraphQL returns empty), so the source of stars can only be inferred from when posts went up.

2. Competitors

Name	Form	Launched	Scale	How it differs from Simlock
EAS Simulator(Expo)	Big tech / commercial cloud	2026-09-14 preview(x.com/expo/status/2099491302056616215)	Expo ecosystem, scale not disclosed	Cloud-hosted, pay-as-you-go, with session replay. Simlock is self-hosted, free, and runs on your own Mac. Julius from T3 Code said: "a great alternative to EAS Simulators if you happen to have some spare [machines]."
BariBariGood/manzanas	Open source	2026-09-01 Show HN (1 points, 0 comments; our hackernews library)	Very small, star count not verified	Nearly the same idea: a Mac daemon, leases, a warm pool, a fleet of Macs, MCP, but iOS only. Launched two months before Simlock, with no distribution.
entropyconquers/simfleet	Open source	2026-09-25	127 stars (github library)	"Control plane for parallel RN work," with a slimmed-down simulator, Metro port leasing, and build caching. macOS and RN only. Launched one month before Simlock.
AgentCloud(Show HN)	Web / cloud	2026-08-27	1 ratings	MCP-first iOS simulator cloud, hosted
MobAI-App/simslim	Open source	2026-07-20	3,499 stars, +7 in 213 days	Complementary, not a competitor: it only lets one Mac run more simulators (by turning off the background daemon). Under the launch tweet, someone asked whether Simlock would integrate SimSlim's tricks.
mobile-next/mobile-mcp	Open source	2025-03	10,317 stars	Lets agents "operate" devices, doesn't handle allocation or isolation
callstack/agent-device	Open source (Callstack's own)	2026-01	5,277 stars	Same company's device-operation layer. Simlock handles allocation, agent-device handles operation; they're one stack.
lycorp-jp/sim-use, joshuayoes/ios-simulator-mcp, conorluddy/ios-simulator-skill	Open source	2025–2026	1.3k–2.2k stars	All of these operate a single simulator and don't manage multiple agents competing for devices
avito-tech/Emcee, plu/pxctest, AppiumTestDistribution, GADS	Open source (older generation)	2016–2022	200–1,000 stars	Built for CI test parallel sharding and device-farm management, not for interactive device leasing to agents
BrowserStack / AWS Device Farm / Firebase Test Lab	Commercial cloud	Already around	Big	Paid cloud real devices, aimed at QA, not local agents

Finding: the specific angle of "renting simulators to parallel agents" already had at least two open-source implementations before Simlock (manzanas and simfleet), plus Expo's commercial version. None of the open-source ones took off. A week after launch, we found no new copycat libraries in our data (checked for repos created after 2026-10-25 with simulator/emulator + agent in the description), no followers yet.

3. Growth tactics (where users come from)

Locked in a popular agent client integration before launch (confirmed): Julius from T3 Code (@jullerino) posted "You can now use Simlock in T3 Code" at 10-29 22:16 UTC, about 12 hours before the official launch. Callstack's @thymikee replied "Thanks for being the earliest adopter! … give it star on GitHub." T3 Code (pingdotgg/t3code, from Theo's company) was at 48k stars in the same period, +16.5k over 7 days (github library). It was at peak attention, so Simlock got a ride on its exposure.
Company and founders amplified the launch in sync (confirmed): the author 10-30 posted a video; the @callstackio official account replied "let's go, self-host everything"; CEO Mike Grabowski (@grabbou, bio reads "codex ambassador @openai", React Foundation) 10-30 and 11-03 reposted it twice (x.com/grabbou/status/2116270418063479257, …/2117627120910242164), and LinkedIn ran it too. Replies include "great launch as always", which suggests the team has a fixed launch playbook and a fixed audience.
The author's personal brand (confirmed): Szymon Chmal wrote Rozenite, Harness, Voltra, and React Native Grab. He is a Microsoft MVP and a Chain React speaker, and gave the talk "Giving agents hands and eyes in React Native" (YouTube). The RN community already had a loyal following for his work.
README asks for stars directly (confirmed): "If you think it's cool, please star it 🌟", and replies also ask people to star it.
Fast iteration, with numbers used for a second round of promotion (confirmed): 4 days, 70+ PRs merged, three versions in a week. 11-03 then posted another round built around "500 stars / 1000 downloads / 90 issues".
HN / Reddit: We found no Simlock post in our hackernews dataset, and Reddit search turned up nothing. There is no evidence that it went through either channel.
Paid acquisition: not applicable, no evidence found.

4. Why it took off

Timing: In the second half of 2026, running multiple agents in parallel with git worktrees became routine (Claude Code, Codex, and T3 Code all promote parallel sessions). The first physical bottleneck mobile developers hit is the simulator: several agents sharing the same devices visible to simctl/adb overwrite installs, erase them, and shut them down. Comments under the launch post state the pain plainly: "Two agents on one sim is how a test taps a button the other agent just deleted"; "500+ stars and 70+ merged PRs in four days tells me this was not a nice-to-have".
Distribution: 368 stars arrived concentrated on launch day. npm was in single digits per day before launch, so the stars came from one coordinated release, not organic search. T3 Code's early integration, the CEO (an OpenAI Codex ambassador) reposting, and the company's official account boosting all landed on the same day.
Clear positioning: the one-liner "parallel coding agents stop fighting over the same simulator" sits right opposite Expo's paid cloud (EAS Simulator, launched in September), pitching "self-hosted, free, use your idle Mac".
Product completeness: CLI + MCP + HTTP API + Node client + web console, supporting iOS/Android/real devices, usable on a single machine or across a fleet, with a complete docs site. That is broader than manzanas or simfleet, which cover only one platform or one framework.
Backed by: Callstack is a core RN contributor, and its org ships a set of device tools including agent-device (5.3k stars). Simlock launched framed as one piece of Callstack's "agent mobile testing infrastructure" (grabbou's 11-03 post: "we're building infrastructure for agents to test apps on your own hardware").

5. Differentiation, and whether it's really why it grew

The differentiation has three parts: (a) self-hosted and free under MIT, versus Expo's paid cloud; (b) one unified "lease" model across iOS, Android, and real devices (queueing, on-demand creation, TTL auto-reclaim, deletion after use), plus a gateway that pools multiple machines into one; (c) every access method (CLI/MCP/HTTP), with T3 Code already integrated. But manzanas, an open-source implementation of the same concept (iOS, multiple Macs, leases, MCP), already existed in September and got only 1 points on HN, and nobody noticed. So differentiation is a necessary condition, not the main cause. The main cause is Callstack's distribution network and T3 Code's launch-day integration. Breadth and completeness determine whether people who arrive will star it and stay to file issues.

6. Visibility

It circulated in small circles: RN/mobile developers and the agent-tooling crowd on X. The author posted star and download counts on X and LinkedIn himself, but there was no press coverage and no HN discussion, and our x-radar corpus has nothing on it (x-radar only covers specific circles). It is a free open-source tool, so there is no question of disclosed revenue.

7. Takeaways

Before launch, a hot agent client (T3 Code) integrated it first, and the partner posted about it the night before, so at launch there was already visible evidence that people were using it.
The team posted together on the same day: the author posted a video, the CEO and the official account relayed it, LinkedIn ran it in parallel, and on day 4 they posted another round with the growth numbers.
They picked a friction point newly created by the agent era (resource contention, isolation, leases), gave it a name that is understood at a glance, and stated the pain in one sentence.
They benchmarked against a paid cloud product (EAS Simulator) and sold "self-hosted, free".
They polished it privately for over 3 months before going public (repo created in July, on npm in August, public at the end of October), so docs, console, and multiple access methods were all ready at launch.

8. Uncertain

How many of the 610 stars are people actually using? npm shows a few hundred daily downloads, which includes CI and mirror traffic. Not verified.
Of "90+ issues / 70+ PRs", how many were filed by the team or by agents? Looking at the commit history, almost all are from the author alone. The share of external contribution was not verified.
The stargazer list is hidden, so we cannot tell which post or which country the stars came from.
T3 Code's integration is still behind a Labs toggle, so how many users it actually brought in is unclear.
Our X dataset does not capture engagement on these Callstack posts. Like and repost counts are unverified.

Sources: GitHub API (2026-11-06), our github/hackernews datasets, the npm registry and download API, X posts (read via r.jina.ai), LinkedIn posts, incubator.callstack.com/simlock docs, expo.dev/services/simulators.

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.