AskAlf – 연중무휴로 운영되는 자체 호스팅 AI 인력

hackernews | | 📦 오픈소스
#ai인력 #askalf #claude #openai #review #생산성 #자동화 #자체호스팅
원문 출처: hackernews · Genesis Park에서 요약 및 분석

요약

AskAlf는 자체 호스팅 가능한 AI 워크포스로, 사용자가 필요로 하는 마케팅, 지원, 개발 등 다양한 분야의 전문가 에이전트를 자동으로 생성하고 24시간 운영합니다. 이 플랫폼은 Docker를 통해 간편히 설치할 수 있으며, 버그 수정이나 보안 취약점 패치와 같은 실제 작업 사례에서 한 달 0.43달러(약 600원)라는 저렴한 비용으로 효율성을 입증했습니다. 또한, 클로드 코드와 오픈AI Codex 터미널을 내장하고 GitHub 및 아마존(AWS) 등 50여 개 이상의 외부 툴과 연동되어 실제 머신과 시스템을 제어할 수 있는 것이 특징입니다.

본문

Tell Alf what you need. Alf builds the team. One AI that creates specialized workers for any task — marketing, support, e-commerce, research, finance, personal productivity, whatever you need. Alf figures out who to hire, puts them to work, and keeps everything running 24/7. Other tools give you a chatbot. AskAlf gives you a workforce. askalf.org · Discord · Discussions · @ask_alf curl -fsSL https://get.askalf.org | bash That's it. Checks prerequisites, generates secrets, pulls images, and starts your team at http://localhost:3001 . Runs on Linux, macOS, and Windows WSL2 — anything with Docker. Manual install git clone https://github.com/askalf/askalf.git cd askalf && ./setup.sh docker compose -f docker-compose.selfhosted.yml up -d Your team runs 24/7. Here's a real cycle: - 10:49 PM — A monitor catches a regression. Creates a ticket. - 10:51 PM — A builder claims it, traces the root cause, writes the fix. 12 minutes later. - 11:03 PM — Tests re-run. All pass. Ticket resolved. - 11:14 PM — Security scans 847 dependencies. Finds 2 CVEs. Patches both. - 01:15 AM — Another worker detects an API contract break from the fix. Updates it. - 02:04 AM — Monitor flags Redis at 91%. Consults memory for past fixes. Applies it. Redis drops to 52%. - 03:12 AM — A Compliance Auditor is spawned on demand to check regulatory drift. Cost: $0.04. - 04:30 AM — Writer generates a status report. - 06:47 AM — Cycle complete. 6 workers, 18 executions, 8 hours. Total cost: $0.43. - Next cycle begins automatically. Alf briefs you whenever you check in. Tell Alf what you need. Alf creates the right specialist — for any industry, any task. - You describe what you need in plain English - Alf figures out what kind of specialist is required - If one exists on your team, Alf assigns them. If not, Alf creates one. - The specialist executes with the right tools, system prompt, and domain knowledge - Results are stored in memory — next time is faster Competitor Researcher · SEO Analyst · Customer Support Agent · Invoice Monitor · Content Writer · Security Scanner · Data Analyst · Compliance Auditor · Social Media Monitor · Review Responder · Report Generator · Meal Planner · Travel Researcher · Habit Tracker · Budget Coach · Pet Care Scheduler · and any specialist you need 109 built-in templates across 16 categories. Workers are first-class — they create tickets, store memories, learn from experience, and coordinate with each other. Browse, share, and install worker templates from the community: - Submit your custom templates to the public library - Install community-created skills with one click - Rate and review templates to surface the best ones - Import/Export skill bundles as JSON — share entire packs with your team - Alf-curated featured skills reviewed and approved by the platform Full Claude Code and OpenAI Codex terminal sessions embedded directly in the dashboard. Not wrappers — real PTY sessions via xterm.js with your full project available. - Claude Code — with MCP tools, your knowledge graph, and full platform context - OpenAI Codex — with dynamic instructions injected from the platform - Switch between them with a tab. Same toolbar. Same workspace. Every task, every outcome, every interaction is stored in a cognitive memory system that grows with every execution. - Semantic Memory — Facts, concepts, system architecture (pgvector embeddings) - Episodic Memory — What happened, what worked, what failed - Procedural Memory — Successful patterns extracted into reusable templates - Knowledge Graph — 1,500+ nodes with automatic cross-agent connection discovery - Metabolic Consolidation — Continuous cycles that strengthen useful memories and decay noise The 10th task is faster than the 1st. The 100th research report knows exactly where to look. Your workers don't just run in the cloud — they control real machines. | Category | Devices | |---|---| | Compute | CLI Agent, Docker Host, SSH Remote, Kubernetes | | Desktop & Mobile | Browser Bridge, Desktop Control, VS Code, Android, iOS | | IoT & Edge | Raspberry Pi, Arduino/ESP32, Home Assistant | Connect any device in one command: npm i -g @askalf/agent askalf-agent connect --url ws://your-server:3005 --name prod-box --install Auto-detects OS, installs as service (systemd/launchd/Windows), runs on boot. OAuth credentials sync automatically — no manual auth on remote devices. - Landing page: askalf.org on Cloudflare Pages (global CDN) - Install script: curl -fsSL https://get.askalf.org | bash on Cloudflare Pages - API: api.askalf.org via Cloudflare Tunnel - Docker Hub: Pre-built images auto-published on every release See @askalf/agent. Slack · Discord · Telegram · WhatsApp · Teams · OpenClaw · REST API · Webhooks · Zapier · n8n · Make · Email · Twilio SMS · SendGrid · Twilio Voice · Zoom | Category | Integrations | |---|---| | Source Control | GitHub · GitLab · Bitbucket | | CRM | Salesforce · HubSpot · Pipedrive | | E-Commerce | Shopify · Stripe · WooCommerce · Square | | Marketing | Mailc

Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.

공유

관련 저널 읽기

전체 보기 →