Mercury

@pengjiayou Its picks list of available models automatically, try: > mercury doctor For more: mercury.cosmicstack.org/docs.html

DeepSeek v4 发布了？装个 Mercury Agent 体验一下。

Open source AI agents need memory, speed, and control. Mercury v1.0.0 delivers all three. Release notes: github.com/cosmicstack-la… GitHub: github.com/cosmicstack-la…

Mercury Agent v1.0.0 is live. Major upgrade shipped. Second Brain adds persistent memory across conversations for users, projects, goals, habits, and context. Better responses, improved Telegram controls, faster CLI streaming, clearer tool feedback, cleaner architecture. Mercury keeps getting sharper.

MERCURY started with a QUESTION: What if Your AI Agent respected Your PERMISSIONS, Your TOKEN BUDGET, and Your IDENTITY PREFERENCES, all at once? Turns out NOBODY had BUILT that yet. So WE Did.

We’re revealing our official logo soon… ✌️

@GithubProjects We’re on a budget ;)

So, how many tokens did you consume today?

@micLivs @Ctrl_Alt_Zaid What made you archive it? It seems great work.

@Ctrl_Alt_Zaid Sorry guys i made it first github.com/Michaelliv/mer…

OpenClaw sparked the Idea. Hermes brought the Energy. Mercury now delivers true CONTROL. Token-efficient by design. Strict budget guardrails. Dangerous commands blocked. Soul files fully editable. Agents that finally behave predictably. This is OpenClaw + Hermes, perfected.

@HuggingModels Thanks for showcasing ❤️

OpenClaw users after knowing about Mercury Agent with token optimisation and permission scoped.

@GithubProjects Thanks for showcasing ❤️

@Ctrl_Alt_Zaid Thanks for such a detailed article ❤️

Mercury is our take on what AI agents should have been from the start. Built with memory, guardrails, token discipline, and real control. Inspired by the progress of tools like OpenClaw and Hermes, then pushed further. Introducing Mercury.