A swarm of AI agents that hacks like a red team, plus the fix for models that can't tell time

The Lens

Time fixes a small but real gap: models do not know what time it is. It exposes two tools, one for the current time in any IANA timezone and one for converting a clock time between zones, and detects your system timezone automatically. Run it with uvx mcp-server-time or Docker. Override the detected zone with --local-timezone if you need to. Nothing to pay. The MCP steering group maintains it as a reference implementation. It does one job: conversion takes an HH:MM time, not a full date, and there is no calendar logic. Cheap to add when deadlines or cross-timezone coordination come up in your prompts.

The Lens

T3MP3ST points a swarm of AI agents at a target and tries to break in, then writes up how. It automates the red-team loop, reconnaissance, exploitation, and reporting, across web apps, CTF challenges, source code, embedded systems, and smart contracts. Instead of asking for its own API keys, it drives coding agents you already run, Claude Code, Codex, or Hermes. AGPL-3.0 and free. Getting it running is light: npm install, npm run server, and a web War Room interface comes up. What is behind that interface is the honest question. The recon operator is production-ready, but the headline, a coordinated multi-agent swarm chaining exploits end to end, is still largely scaffolded. The benchmark numbers (around 90 percent on a web-security challenge suite, most of a held-out CVE set matched) come from a single agent, not the swarm the name promises. To its credit, the repo ships a verify-claims command so you can re-derive them yourself. This is for security people doing authorized work, pentesters, CTF players, researchers probing their own code, not a push-button tool for anything you do not own. Solo researchers and small security teams: worth a look as an AI-assisted recon and triage layer today, with the exploitation swarm as a promising work in progress. Treat the automation as a force multiplier for a human who knows what they are doing, not a replacement for one. The catch is the gap between the pitch and the product. The recon works, the autonomous kill chain does not fully exist yet, and AGPL-3.0 means anything you build on top inherits that copyleft. Judge it on what runs today, not the roadmap in the README.

The Lens

SigNoz puts traces, metrics, and logs in one dashboard, the three pillars of observability you'd otherwise stitch together from separate tools. It's built on OpenTelemetry, the open standard for collecting this data, so your instrumentation isn't locked to a proprietary SDK and works with any compatible backend. The self-hosted version is free and holds nothing back: same codebase as the cloud, every feature included. Running it yourself means running ClickHouse, which is the real ops consideration. ClickHouse is fast but hungry, so plan for a VM with 8GB+ of RAM and storage that grows with your data. Docker or Kubernetes, a couple of hours to stand up. If you have the ops capacity, this is where the savings live. Solo developers: the cloud free tier or a small self-hosted box is plenty. Small teams with ops capacity: self-host. Everyone else: SigNoz Cloud starts at $49/mo for the Teams plan plus usage ($0.30/GB for logs and traces, $0.10 per million metric samples), and it still undercuts a Datadog bill many times over. That's the whole pitch, Datadog-level visibility without the Datadog invoice. The catch is maturity. SigNoz is younger than the Grafana and Prometheus stack, and its plugin and integration ecosystem is smaller. If you need hundreds of prebuilt dashboards and integrations, Grafana still wins. If you want one clean unified platform instead of three tools duct-taped together, SigNoz is the better shape.

The Lens

Home Assistant runs your smart home on your own hardware instead of somebody else's cloud. Lights, thermostats, cameras, locks, sensors, media players, vacuums: one dashboard and one automation engine sitting on top of more than a thousand built-in integrations covering thousands of device brands. Apache 2.0, free forever, no feature tiers. Everything stays local by default, which is the actual reason to run it. Your motion sensor data does not leave the house unless you send it somewhere. Install Home Assistant OS on a dedicated box and it manages its own updates and add-ons; run the Docker container instead and you own the update cycle. Either way, plan on a weekend for the first setup. YAML is mostly gone from daily use but still turns up in advanced automations. Hardware is where the math changed in 2026. Home Assistant Green is $199 with the OS preinstalled, plus $49 for the Connect ZBT-2 dongle if you want Zigbee or Thread. Building on a Raspberry Pi 5 used to undercut that easily, but two memory-driven price hikes this year pushed the 16GB board to $305, so official hardware is now the cheaper path unless you already own a spare machine. Home Assistant Cloud (Nabu Casa) stays $6.50/mo or $65/yr for remote access, voice assistants, and automatic SSL, and it funds the project. The catch: integrations break. Releases ship monthly and a vendor API change can take a device offline until somebody patches it. 2026.8 alone dropped Volvo On Call and Permobil and forced a UniFi Protect upgrade. This maintains itself about as well as any hobby does.