heggria

heggria / Hermit

Public

Local-first governed agent kernel for durable, inspectable, operator-trust-oriented AI work.

heggria.github.ioHermit agent ai ai-agent approvals automation
21
4
100% credibility
Found Mar 18, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Hermit is a local-first AI agent framework that provides governed execution with human approvals before actions, inspectable receipts and proofs afterward, and rollback support for safe, auditable workflows.

How It Works

1
🔍 Discover Hermit

You hear about Hermit, a safe local helper that lets AI do tasks while keeping everything visible and reversible.

2
📥 Install easily

Run one simple command on your Mac and it sets up your personal AI space, grabbing any existing AI logins automatically.

3
🔗 Connect your AI

Link your favorite AI like Claude or ChatGPT – it pulls in your saved details without hassle.

4
🚀 Run your first task

Tell it something like 'summarize this folder' and watch it work with your approval before making changes.

5
📋 Review what happened

Check the task list, see proofs of what it did, and even roll back if needed – everything stays recorded.

6
☰ Keep it handy

Use the menu bar app to start ongoing help, switch settings, or peek at logs anytime.

✅ Safe, smart assistance

Now you have a trustworthy AI partner for daily tasks, always inspectable and under your control.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 21 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is Hermit?

Hermit is a Python-based local-first agent kernel that runs AI workflows with built-in governance, keeping everything inspectable and recoverable on your machine. You give it tasks via CLI like `hermit run "Summarize this repo"`, and it proposes actions for approval before any side effects, then generates receipts, proofs, and task ledgers you can query with `hermit task list` or `hermit task proof`. It solves the black-box problem of typical agents by prioritizing operator trust over blind execution, with macOS menubar support for quick access.

Why is it gaining traction?

Unlike chatty agents that vanish into logs, Hermit leaves durable records—task states, approvals, and rollbacks via `hermit task rollback`—making audits trivial without cloud dependency. Its local-first design fits the agent trend on GitHub, emphasizing verifiable steps over "helpful now" speed. Developers dig the policy gates that pause for human nods on risky moves, plus seamless imports from tools like Claude or Codex.

Who should use this?

AI tinkerers building personal coding agents that need to ask before editing files, or ops folks running scheduled local tasks worth auditing later. Ideal for solo devs wanting interruptible, stateful assistants without vendor lock-in, like a Feishu-integrated bot for team workflows. Skip if you just need quick chats—Hermit's for trust-heavy, long-running agent work.

Verdict

Early alpha (21 stars, 1.0% credibility) with solid docs and CLI, but test it via one-command macOS install before production. Worth watching for local-first agent fans; prototype your governed workflows today.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.