3DCF-Labs

3DCF-Labs / safepilot

Public

AI assistant that executes real work, safely.

ai assistant own-your-data personal safe
378
9
100% credibility
Found Feb 26, 2026 at 345 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

SafePilot is a self-hosted Telegram bot powered by AI that plans and executes tasks like code changes, searches, and integrations with services such as GitHub, Slack, and Jira, featuring safety approvals and persistent workspaces.

How It Works

1
📱 Discover SafePilot

You hear about SafePilot, a friendly Telegram bot that helps automate real work safely using AI.

2
👋 Chat with your bot

Add the bot to Telegram and say hello to get started with simple setup.

3
🏠 Create your workspace

Pick a role like general helper or developer assistant, and set safety preferences to match your needs.

4
🔌 Connect services

Link tools like email, calendars, or code repos so the bot can help across your apps.

5
💬 Send a task

Tell the bot what you want done, like 'check my tasks' or 'update the code'.

6
✅ Review and approve

See the plan, approve safe steps instantly, and confirm riskier ones for peace of mind.

🚀 Work gets done

Watch tasks complete automatically, get summaries, and free up time for what matters.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 345 to 378 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 safepilot?

SafePilot is a Rust-powered, self-hosted AI assistant accessed via Telegram that converts chat messages into safe, executable task plans across integrations like GitHub, Slack, Notion, Linear, Jira, Todoist, and shell commands. It persists runs in SQLite, compresses context for long sessions, and executes jobs in isolated workspaces with built-in safeguards like task approvals and SSRF protection. Developers get hands-off automation for real work—cloning repos, running tests, posting updates—without exposing systems to unchecked AI.

Why is it gaining traction?

Its defense-in-depth security stands out: risky actions like shell runs or writes halt for explicit approval via /approve or temporary /unsafe mode, unlike looser assistants on GitHub such as home assistant integrations or CISO tools. Docker Compose deploys harden it out-of-box with non-root users and read-only filesystems, hooking devs tired of prompt injection risks in agents like Bytedance's Doubao. With 368 stars, it pulls users seeking production-ready task execution over toy prototypes.

Who should use this?

DevOps engineers chaining GitHub CI checks with Jira tickets, indie hackers querying weather or searching via Brave before Notion updates, or teams binding public Telegram channels to scoped runtimes for Slack announcements. Ideal for backend devs automating code reviews or Linear triage without vendor lock-in.

Verdict

Try it if security trumps speed—solid docs, CI badges, and Apache license make setup straightforward, but 1.0% credibility and modest stars signal early maturity; test in Docker first before prod. Promising for safe AI ops, beats home assistant GitHub clones for task-heavy flows.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.