2026-04-11

• #0002

  #0002Agents & toolsHacker News · Show HN AIlast month

  `Eve`: Managed `OpenClaw` for Work

  60radar

  EveAgent service — sandbox, browser, and connectors unified

  A background agent runner bundles a real Linux sandbox, browser, code execution, and 1000+ connectors behind a hosted UI. It is ambitious on orchestration and async workflows, but trust, reliability, and pricing discipline will decide whether it sticks.

  • —Each task runs in an isolated Linux box with 2 vCPUs, 4GB RAM, 10GB disk, plus a real filesystem and headless Chromium; that is enough for many practical automations.
  • —An orchestrator on Claude Opus 4.6 routes subtasks across coding, browsing, research, and media generation; the product pitch is not a single model, but a managed multi-agent stack.
  • —Parallel sub-agents coordinate through a shared filesystem and keep persistent memory across sessions; repeated workflows can compound context instead of restarting cold every time.
  • —The surface area is unusually broad: web UI for live execution, iMessage for async task dispatch, and preinstalled role skills across sales, marketing, and finance.
  • —The $100 free credits lower trial friction, but high-risk claims like tax filing push the product into a trust-sensitive zone where errors get expensive fast.

  Source: eve.new/loginRead original →
• #0001

  #0001Agents & toolsHacker News · Show HN AIlast month

  `Twill.ai`: Run coding CLIs in cloud sandboxes and get back PRs

  50radar

  Twill.aiCloud coding agent ops — per-task sandbox isolation

  Moves coding agents off the laptop into per-task cloud sandboxes, then returns PRs, reviews, or diagnoses through Slack, GitHub, or Linear. Useful if you want overnight parallel work without handing local machine access to agents, but the moat lives in orchestration quality.

  • —Each task gets its own sandbox, repo clone, dependency install, and chosen CLI run, with filesystem, port, and process isolation built in.
  • —Instead of replacing Claude Code or Codex, it wraps lab-native CLIs in the cloud. That keeps model choice flexible and improves automatically as upstream CLIs get better.
  • —Completed runs snapshot the sandbox so the next task on the same repo starts warm. That directly targets the biggest drag in cloud-agent workflows: cold-start setup time.
  • —Inputs can come from Slack, GitHub, Linear, web app, or CLI, and outputs can be PRs, reviews, diagnoses, or follow-up questions. This is closer to agent ops than an IDE addon.

  Source: twill.aiRead original →