@DanKornas: Let AI help administer Linux — without handing it an unrestricted shell. SysKnife is a Linux system administration MCP …

Let AI help administer Linux — without handing it an unrestricted shell.

SysKnife is a Linux system administration MCP server and CLI for builders who want AI assistance with system tasks.

It helps you turn plain-language requests into reviewable plans by having the model select typed actions while a separate daemon builds and executes the commands. You can inspect each step’s risk level before approving it.

Key features:

• Typed actions – the model selects defined operations instead of supplying arbitrary shell commands • Terminal approvals – MCP execution requires one-time receipts issued through a separate terminal approval step • Verifiable audit trail – an Ed25519-signed, hash-chained log records executions and supports integrity checks • MCP integrations – use the planning and execution tools from Claude Code, Cursor, or Codex CLI • Local or hosted models – configure Ollama or providers including OpenAI, Anthropic, and Gemini for planning

It’s open-source (MIT license).

Link in the reply 👇

🔗 GitHub: https://github.com/lacs-project/sysknife…

✉️ If you’re into AI, ML, agents, and building real systems, join my newsletter (it’s free): https://dankornas.substack.com

Your agent needs more than a chat window to manage ongoing work.

Octopus Superintelligence OS is a research-preview agent execution system for builders experimenting with persistent tasks, project permissions, and extensible tools.

It helps you organize agent work beyond individual prompts by giving projects their own context, agents defined roles, and tasks you can return to later.

Key features:

• Project boundaries – give each goal its own context and permissions rather than sharing access across projects • Resumable tasks – leave work running, come back later, and pick up where it stopped • Desktop and terminal clients – access the same Home, Projects, and Agents from either interface • Extension folders – add providers, channels, workers, tools, and checks through JSON descriptors plus scripts or binaries where needed • Workspace drivers – connect work to folders, git worktrees, browsers, containers, or remote machines

The README’s setup requires macOS 14+, Node 24, stable Rust, and a ChatGPT account. Self-improvement is a research aim, not a demonstrated result.

It’s open-source (MIT license).

Link in the reply 👇