Argusic tests whether real software actually runs 👋 Every developer knows this moment. The README says "just npm install". Two hours later you're deep in dependency conflicts, missing env vars and a model that needs a GPU nobody mentioned. Argusic exists to answer one question with evidence: Does it actually run? 🤖 What Argusic does AI agents take real software (open source repos and Hugging Face models) and install, launch and debug it exactly like a new engineer on day one. Clean, single-use machine. Pinned commit. No human help. 📊 So far: 3,163 projects tested, every run recorded. 🔍 How it works 1️⃣ Every project is pinned to an exact commit 2️⃣ An agent gets one mission on a disposable machine: Make it work 3️⃣ A failure earns a second attempt 4️⃣ Two failures settle "could not verify". Never "broken". We don't blame code we couldn't run. 🧮 How we score (out of 100) Install 30, launch or tests 30, error recovery 20, verification depth 20. A deterministic program does the scoring. No AI model grades anything, ever. "Verified" means the agent proved the output works, not just that a process started. 🧾 Proof over opinion Every published number links to its full log and terminal recording. Nothing on Argusic is a review, and no one can pay to change a result. Doubt a verdict? Open the log and check it yourself. 🔬 Methodology: Full rules, scoring and changelog argusic.com/methodology 🏆 Best software, tested: Databases, CMS, AI agent frameworks and more, ranked by whether they actually install and launch argusic.com/best 📰 Reports: Daily and weekly summaries of what ran, what failed and what it cost argusic.com/reports 🖥️ Servers: 333 GPU cloud and VPS offers from 114 providers, with verified prices and measured spin-up times argusic.com/servers 💻 Hardware: 1,300+ laptops, desktops and mini PCs for programming and AI, with estimates of which local models each can run argusic.com/hardware 🧪 Maintainer or builder? Request a test for your repo argusic.com/request 📂 Open data, free under CC BY 4.0, updated daily. Use it, cite it, build on it: 🐙 GitHub: Every verdict, score and pinned environment github.com/argusic/does-it-r… 🤗 Hugging Face: Load the dataset straight into your notebook huggingface.co/datasets/argu… 📺 Every run on video youtube.com/@argusiccom 🤝 Follow along We build Argusic in public. Follow for daily verdicts, new reports and the occasional install that fights back. Reply with a repo you want us to test next 👇 #Argusic

vector-graph-rag: Graph RAG with pure vector search, achieving SOTA performance in multi-hop reasoning scenarios. zilliztech/vector-graph-rag · Python · MIT · 262 stars vector-graph-rag tested: it runs, but only against stand-in services, never the real thing. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 656bdf2. Install to first launch: 1 min Whole session: 12 min Errors met: 0, fixed without help: 0 Argusic Score: 92/100 Depth: checked against stand-in services only The agent's own closing words: "The full test suite (143 tests across 11 files) passes cleanly and the FastAPI server starts and responds correctly on /health." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #VectorGraphRag #GraphRag #Graphrag #Jev #KnowledgeBase #KnowledgeGraph #Milvus #Argusic

notion-electron: Notion Desktop Client for Linux. anechunaev/notion-electron · TypeScript · MIT · 264 stars notion-electron tested: 3 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 92918eb. Install to first launch: 7 min Whole session: 7 min Errors met: 3, fixed without help: 3 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "The repository installs (npm install), builds (npm run build emits complete main/preload/renderer output), typechecks (tsc --noEmit passes all three tsconfigs), lints (ESLint + Prettier pass), and the Electron application launches and runs persistently on a virtual X display (state S, killed by timeout after 20s) with only expected container-related D-Bus warnings." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #NotionElectron #Electron #Javascript #Nodejs #Notion #Office #Productivity #Argusic

client: Tetiva — desktop API client for HTTP, gRPC, GraphQL and WebSocket (Wails v3 + Vue 3). tetiva-app/client · Go · MIT · 267 stars client tested: 4 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 3eb9e14. Install to first launch: 12 min Whole session: 16 min Errors met: 4, fixed without help: 4 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "Server binary builds and serves HTTP 200 on port 9246; 38 Go test packages pass (36 ok, 2 no test files); frontend dist embedded and served correctly" The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Client #ApiClient #DesktopApp #DeveloperTools #Go #Golang #Graphql #Argusic

animaworks: Organization-as-Code for autonomous AI agents. xuiltul/animaworks · Python · Apache-2.0 · 268 stars animaworks tested: 5 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit f700ff9. Install to first launch: 8 min Whole session: 44 min Errors met: 5, fixed without help: 5 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "17092 unit tests pass; CLI prints full help; demo server binds port 18501 with Uvicorn; all 5 pre-existing test failures fixed via IPv6 loopback detection, zstd Python fallback, and Node 18 ESM compatibility." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Animaworks #AgentFramework #AiAgents #AutonomousAgents #BrainInspired #Claude #Forgetting #Argusic

heimdall: A cloud native Identity Aware Proxy and Access Control Decision service. dadrus/heimdall · Go · Apache-2.0 · 260 stars heimdall tested: 2 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 4e12733. Install to first launch: 4 min Whole session: 32 min Errors met: 2, fixed without help: 2 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "Heimdall builds from source, the binary starts in decision mode listening on configured ports, the management API responds HTTP 200 on /.well-known/health and /.well-known/jwks, and all 101 Go test packages pass with serial execution to avoid the container's inotify instance limit." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Heimdall #AccessControl #AccessManagement #ApiGateway #AuthApi #AuthProxy #Authentication #Argusic

LinkBreeze: The self-hosted Linktree alternative. OmniRise/LinkBreeze · TypeScript · MIT · 260 stars LinkBreeze tested: 5 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit c41555a. Install to first launch: 26 min Whole session: 31 min Errors met: 5, fixed without help: 5 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "LinkBreeze 1.4.1 builds from source, passes all 1064 unit tests, and serves the setup page on port 3344 with health endpoint responding 200." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #LinkBreeze #Analytics #CreatorTool #Docker #DockerImage #LinkInBio #LinkInBioPage #Argusic

workflow: The type-safe, event-driven workflow orchestration library that scales with your business. luno/workflow · Go · BSD-3-Clause · 264 stars workflow tested: nothing broke, ran first time in 2 minutes. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 13d66ce. Install to first launch: 2 min Whole session: 4 min Errors met: 0, fixed without help: 0 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "The repository builds from source with Go 1.26.10, all internal unit tests pass (14 packages, 0 failures), and three in-memory example workflows (gettingstarted, callback, timeout, schedule, connector) execute end-to-end successfully. The embedded Web UI server starts, serves the homepage, and responds to all 3 API endpoints." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Workflow #Durable #Eda #Eventdrivenarchitecture #Go #Golang #Hacktoberfest #Argusic

Argusic daily report, 27 October 2026 (UTC) On 27 October an AI agent tried to install and run 62 open-source projects on clean machines, with nobody helping: 73 attempts in all, 5 of them cut short by our own tooling and not counted. Runs: 44 Runs with mocks: 14 Could not verify: 3 Not yet verified: 1 Five from the day: • AutoSwarm-Shadow-Adapter (3,986 stars): Runs argusic.com/subject/autoswar… • dsh-context (3,719 stars): Runs argusic.com/subject/dsh-cont… • dsh-purge (3,047 stars): Runs with mocks argusic.com/subject/dsh-purg… • mcd-saver-strategist (1,545 stars): Runs argusic.com/subject/mcd-save… • jelly-app (271 stars): Could not verify argusic.com/subject/jelly-ap… All 62 projects, each with its log and recording: argusic.com/reports/daily/20… How it is measured: argusic.com/methodology Every project tested so far: argusic.com/subjects GPU cloud and VPS prices: argusic.com/servers Laptops and desktops for programming and AI: argusic.com/hardware

Ecommerce-Laravel-Bootstrap: Responsive, Multi-Vendor, MultiLanguage Online Store Platform (shopping cart solution). kirilkirkov/Ecommerce-Laravel-Bootstrap · JavaScript · MIT · 264 stars Ecommerce-Laravel-Bootstrap tested: 4 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 0dce561. Install to first launch: 18 min Whole session: 19 min Errors met: 4, fixed without help: 4 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "Installed Laravel 7.x ecommerce app running on PHP 8.1.34 with SQLite, composer dependencies installed, database migrated and seeded with 1 admin user, PHPUnit test suite passing, and web server serving all public pages at 200." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #EcommerceLaravelBootstrap #Administration #Bootstrap #ECommerce #ECommerceplatform #Ecommerce #EcommercePlatform #Argusic

advai-cli: Local-first CLI for agent skills, browser automation, knowledge bases, external CLIs, and terminal AI chat. advai-x/advai-cli · Python · MIT · 265 stars advai-cli tested: nothing broke, ran first time. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit b6ecbce. Whole session: 1 min Errors met: 0, fixed without help: 0 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "advai-cli 1.0.12 installed from source, all 78 unit tests pass, the CLI responds to all top-level commands, and the knowledge base subsystem creates, adds, searches, and syncs real local files." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #AdvaiCli #AgentSkills #AiAgents #Automation #BrowserAutomation #Cli #DeveloperTools #Argusic

biscuit: biscuit is a native, extensible, AI code editor. tomlin7/biscuit · Python · MIT · 266 stars biscuit tested: 5 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 85c5beb. Install to first launch: 12 min Whole session: 12 min Errors met: 5, fixed without help: 5 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "The Biscuit code editor (v3.5.0) installs from source, all 312 tests pass, the CLI responds to --version and --help, and the GUI launches on the virtual display creating the full editor window tree including menu bar, status bar, editor tabs, panels, and sidebars." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Biscuit #Agents #Ai #CodeEditor #Copilot #Cursor #Debugger #Argusic

ai-gateway: Unified AI Gateway for 30+ LLMs (OpenAI, Anthropic, Bedrock, Azure etc) with Caching, Guardrails, A/B test & cost controls. ferro-labs/ai-gateway · Go · Apache-2.0 · 271 stars ai-gateway tested: 3 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit d025ca1. Install to first launch: 4 min Whole session: 10 min Errors met: 3, fixed without help: 3 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "Go 1.25.14 installed from tarball, ferrogw binary built at 42 MB binary, server starts and responds 200 on /livez, /readyz, and /health, 'go test' over all 89 packages passes with zero failures, 'ferrogw validate' confirms config valid." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #AiGateway #AiInfrastructure #Guardrails #Litellm #Llm #LlmCost #LlmGateway #Argusic

Lampline: Build AI phone agents on your own computer and test every change with an AI caller. C4rohan/Lampline · Python · Apache-2.0 · 379 stars Lampline tested: 1 error, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 004aa71. Install to first launch: 5 min Whole session: 23 min Errors met: 1, fixed without help: 1 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "Both packages (vpd-vk 0.3.0, vpengine 0.3.0) installed in .venv-engine; 453 tests pass; engine starts and answers health and chat session endpoints." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Lampline #AgentsAsCode #ContactCenter #ConversationalAi #CxAgentStudio #DialogflowCx #Ivr #Argusic

synaplan: Open-source, self-hosted AI platform (PHP/Symfony): chat with any model, RAG over your docs, embeddable widget, MCP — see the cost of every answer. metadist/synaplan · PHP · Apache-2.0 · 299 stars synaplan tested: 3 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 619349a. Install to first launch: 18 min Whole session: 42 min Errors met: 3, fixed without help: 3 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "Frontend vitest unit test suite passes 3093/3095 tests (99.93%), 435/436 test files pass, with the remaining 2 failed tests being pre-existing ImpersonationBanner router navigation timing issues." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Synaplan #Ai #AiAgents #Anthropic #Chat #ChatbotWidget #ChatgptAlternative #Argusic

uteke: 🧠 The Brain for Your AI — Local-first memory engine for AI agents. codecoradev/uteke · Rust · Apache-2.0 · 277 stars uteke tested: 2 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 7632a85. Install to first launch: 13 min Whole session: 14 min Errors met: 2, fixed without help: 2 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "Uteke v0.18.2 builds from source, stores and semantically recalls memories via CLI (remember/recall), serves HTTP API on port 18768 (health/remember/recall all 200 OK with correct JSON), and the MCP server responds to initialize — all 705 workspace tests pass with 0 failures." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Uteke #AgentMemory #Ai #AiAgents #Cli #CliTool #Embeddings #Argusic

OSMExpress: Fast database file format for OpenStreetMap. bdon/OSMExpress · C++ · BSD-2-Clause · 274 stars OSMExpress tested: 3 errors, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 045a515. Install to first launch: 10 min Whole session: 14 min Errors met: 3, fixed without help: 3 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "osmx C++ binary builds, runs expand/query/extract on real OSM data, test suite passes (27/27), Python osmx package installs and reads databases correctly." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #OSMExpress #Database #Openstreetmap #Spatial #GitHub #Argusic

Sylius-Standard: Sylius app skeleton. Sylius/Sylius-Standard · PHP · MIT · 278 stars Sylius-Standard tested: 7 errors, 6 fixed alone, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 6b58030. Install to first launch: 10 min Whole session: 14 min Errors met: 7, fixed without help: 6 Argusic Score: 97.14/100 Depth: checked against the real service The agent's own closing words: "The Sylius PHP backend boots successfully, its Symfony console commands work, its API responds correctly (200 on /api/v2/shop/products and /api/v2/docs), and its test suite passes; the frontend build failed because Node.js 18 does not meet Sylius's Node.js ^22 requirement, and the shop homepage errors because seed data is missing (no MySQL available to run the full install)." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #SyliusStandard #Ecommerce #EcommerceFramework #EcommercePlatform #Hacktoberfest #Php #Sylius #Argusic

antirion: ⚡ The fastest open-source AI gateway — one Rust binary with a built-in console to route, secure, budget and observe every LLM call across OpenAI, Anthropic, Gemini, Bedrock and more. eliasbiondo/antirion · Rust · Apache-2.0 · 334 stars antirion tested: nothing broke, ran first time in 3 minutes. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 08ad3b6. Install to first launch: 3 min Whole session: 16 min Errors met: 0, fixed without help: 0 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "The Antirion AI gateway builds from source, its test suite passes 1044/1044 tests, the binary starts in standalone mode with the embedded store, and successfully routes a chat completion request through a provider, deployment, model, and key to a mock upstream and back with HTTP 200." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #Antirion #A2a #AiGateway #Anthropic #ApiGateway #BifrostAlternative #Guardrails #Argusic

dataCompare: big data comparison and data profiling platform: low code，data comparison and data profiling. dromara/dataCompare · Java · Apache-2.0 · 280 stars dataCompare tested: it runs, but only against stand-in services, never the real thing. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit 906df90. Install to first launch: 14 min Whole session: 24 min Errors met: 3, fixed without help: 3 Argusic Score: 92/100 Depth: checked against stand-in services only The agent's own closing words: "Spring Boot dataCompare app runs on port 80, connects to local MariaDB database with 36 tables, serves login page, and authenticates with admin/admin123 credentials." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #dataCompare #Bigdata #Doris #Hive #LowCode #Spark #Java #Argusic

agent-chrome-relay: Let your AI agents use your Chrome, signed in as you, in background tabs. aindeev/agent-chrome-relay · Go · MIT · 282 stars agent-chrome-relay tested: 1 error, no human help, still ran. It was installed and run by an Argusic agent on a clean machine, with no human help, from the repository pinned at commit ca2f952. Install to first launch: 19 min Whole session: 20 min Errors met: 1, fixed without help: 1 Argusic Score: 100/100 Depth: checked against the real service The agent's own closing words: "The Chrome Relay relay server starts, answers HTTP endpoints with correct status codes, and passes all 12 integration tests including agent WebSocket connections authenticated by secret and process markers, extension origin validation, web page rejection, and audit trail persistence." The whole session is in this video, unedited, at the speed it actually happened. Every number above is read from that run's own log. Follow for a new repository put through this every day. #AgentChromeRelay #AgentBrowser #AiAgents #BrowserAutomation #Cdp #ChromeExtension #Claude #Argusic