@GithubProjects Our GitHub Repository:
github.com/cosmicstack-la…
English
Zaid
270 posts
Zaid
@Ctrl_Alt_Zaid
Non-Technical Member of Technical Staff...
شامل ہوئے Haziran 2025
7 فالونگ390 فالوورز
|￣￣￣￣￣￣￣￣￣￣￣￣￣|
| Share your GitHub profile. |
|＿＿＿＿＿＿＿＿＿＿＿＿＿|
\ (•◡•) /
\ /
——
| |
|_ |_
English
Join us to make something better for a better future.
GitHub Repository:
github.com/cosmicstack-la…
English
We shipped Mercury Agent v1.0.0.
Second Brain adds persistent memory. Responses are sharper, streaming is faster, and the overall experience feels cleaner.
Mercury@mercury__agent
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.
English
@fridgebuzz_art Separate projects can make more sense than forcing different priorities into one agent.
Self-improvement and security aren’t opposites either. Different tools can optimize for different jobs.
English
@Ctrl_Alt_Zaid It would have been nice to see these improvements added to Hermes rather than making a whole different agent. Now you have to give up the self-improving agent for added security features. Not ideal.
English
@4egui @GithubProjects open-source builds by iteration.
Ideas spread, implementations compete, users decide.
English
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.
Zaid@Ctrl_Alt_Zaid
English
@luckyPipewrench @GithubProjects Once agents hit real tools, self-policing isn’t enough.
That’s why Mercury uses permissions, constrained execution, and human-controlled workflows.
External controls matter. Internal guardrails do too. Both have a place.
English
Control gets real fuzzy the second the agent can hit tools, browsers, and APIs on its own.
OpenClaw, Hermes, Mercury, all of them run into the same security problem there. If the agent is allowed to police itself, you are trusting the thing under attack to explain why its own actions were fine.
Pipelock puts that control outside the agent and watches the actual traffic.
English
Thanks for sharing and quoting our article, we really appreciate it ❤️
Glad you highlighted Mercury Agent and its focus on reliable control through token budgets, hardened permissions, and those fully editable soul files.
If you're building with it or have any suggestions, pull requests are always welcome on the repo.
English
@Ctrl_Alt_Zaid @grok It gets promoted to Super.
From Start to everything.
English
Hey @grok, what does the Windows key do on Linux… besides having an identity crisis?
GitHub Projects Community@GithubProjects
Hey @grok, what does this key do when you install Linux on a Windows laptop?
English
@GithubProjects @grok It creates a meme on Microsoft and posts it online. 😉
English
