Mercury Cloud Gives Open Agents A Cockpit
Mercury Cloud adds remote control, chat, approvals, memory sync, model switching, and live status to the open-source Mercury Agent while keeping its runtime on the developer’s machine. The service starts free with one agent connection and free-enabled models, with paid tiers beginning at $19/month.
Mercury Cloud targets a real gap in local-first agents: keeping them accessible without exposing ports or maintaining a custom relay. The developer-owned runtime is compelling, but hosted authentication, model budgets, and cloud memory still create a meaningful trust boundary.
- –One command pairs the local agent through a persistent WebSocket connection, avoiding port forwarding, reverse proxies, and certificates
- –The browser cockpit can send messages, approve actions, install skills, inspect memory, switch models, and monitor agent health
- –Mercury Agent remains open source under the MIT license, while Mercury Cloud is the hosted control plane built by Cosmic Stack
- –Free access lowers the barrier to experimentation, while $19/month Pro and higher tiers monetize managed model usage and expanded workflows
- –Shared memory across agents could become a differentiator, but developers should evaluate what context leaves their machine
DISCOVERED
2h ago
2026-08-20
PUBLISHED
2h ago
2026-08-20
RELEVANCE
AUTHOR
mercury__agent