v2.6.0 · OPEN SOURCE · SELF-HOSTED · DAILY UPDATES

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. Self-hosted. Open source.

Other tools give you a chatbot. AskAlf gives you a workforce. Runs on anything with Docker.

Deploy Now Star on GitHub
$ terminal
# One line. That's it.
curl -fsSL https://get.askalf.org | bash

# Checks Docker, generates secrets, pulls images,
# starts your team at http://localhost:3001
☁️
Hosted Version
Coming Soon
🖥️
Live Demo
Coming Soon
🦙
Ollama Support
DeepSeek R1 Llama 3.2 Qwen 3 Gemma 3 Phi 4 Mistral
Run Local · $0 Cost
WHO IT'S FOR

Pick your story.

Same platform. Different problems. Click a tab to see what AskAlf does for you.

You tell Alf what you need before bed. By morning, it's done.
09:15 PM
Travel Planner
You say: "Plan a trip to Austin next month." Alf spawns a Travel Planner. Researches flights, hotels, restaurants. Full itinerary by morning.
09:22 PM
Budget Coach
Analyzes your subscriptions. Finds 3 services you haven't used in months. $47/month in savings.
10:00 PM
Meal Planner
Generates next week's meal plan based on your preferences. Shopping list ready. Accounts for leftovers.
07:00 AM
News Curator
Morning digest: 12 articles from your interests, zero noise. Competitors, industry trends, and one article that changes your mind about something.
$0.18. Trip planned. Budget optimized. Meals sorted. News curated. You slept through all of it.
Your competitors launched a new feature. By morning, you have the analysis, the counter-strategy, and the ad copy.
08:30 PM
Competitor Tracker
Detects your competitor updated their pricing page. Captures changes, compares to last snapshot. Full diff report.
09:15 PM
SEO Analyst
Runs keyword gap analysis against competitor. Finds 14 high-intent keywords they rank for that you don't. Content briefs generated.
11:00 PM
Ad Copy Generator
Drafts 6 ad variations for Google and Facebook targeting the competitor's weakness. A/B test ready.
06:00 AM
Social Monitor
Scanned Twitter, Reddit, LinkedIn for brand mentions. 3 positive, 1 needs response. Draft reply waiting.
$0.37. Competitor analyzed. SEO gaps found. Ads drafted. Social monitored. Ready whenever you are.
Your team pushes code at 10:47 PM. By 9 AM, bugs are fixed, vulnerabilities are patched, and infrastructure is stable.
10:49 PM
Test Runner
Last commit touched the checkout flow. Runs 42 regression tests. 3 fail. Ticket created, assigned to a builder. Nobody wakes up.
10:51 PM
Builder
Traces it to a race condition on line 412. Writes the fix, commits, tests re-validate. 12 minutes after the team went home.
11:14 PM
Security Scanner
Scans 847 packages. Finds CVE-2026-1847 in the payment library. Critical. Builder patches it before midnight.
02:04 AM
System Monitor
Redis at 91%. Consults Alf's memory for past fixes, applies the solution. Redis drops to 52%. Zero downtime.
$0.43. 2 bugs fixed. 1 CVE patched. 1 outage prevented. Your team slept through all of it.
Your Shopify store runs 24/7. So does Alf. Reviews, inventory, sales analysis — handled before you open your laptop.
11:32 PM
Review Monitor
New 1-star review on your bestseller: "Arrived damaged, no response from support." Creates a ticket for Support with order details pulled from Shopify. Flags it as urgent.
01:15 AM
Inventory Tracker
Your top SKU dropped below reorder threshold. Tracker creates a purchase order draft, notifies your supplier channel, and updates your inventory spreadsheet. You won't stock out.
06:00 AM
Sales Analyst
Weekly report generated: revenue up 12%, cart abandonment spiked on mobile, discount code SPRING20 drove 34% of conversions. Formatted and waiting in Slack.
08:00 AM
Support Agent
Drafts a personalized response to the 1-star review with a replacement offer. Queued for your approval. Response time: under 9 hours instead of 2 days.
Reviews caught. Inventory managed. Sales analyzed. ~$0.38/day. Your store runs 24/7 — and now so does your team.
You manage 8 client projects. Each one gets its own Alf team. Reports, monitoring, and maintenance — running continuously across all 8.
Client A
Monitor
Their landing page went down at 1 AM. Monitor detected it, restarted the container, posted in your Slack channel. Client never knew. Uptime: 99.98%.
Client B
Tester
Their developer pushed a breaking change. Tester caught it in staging before production. Ticket created, fix deployed. Client's users never saw the bug.
Client C
Security
Their WordPress plugin has a known XSS vulnerability. Security finds it, patches it, verifies the fix. You invoice them for "proactive security maintenance." They love you for it.
Monday 9 AM
Writer
Status reports generated for all 8 projects automatically. Formatted, filed, ready to forward. Your Monday standup takes 5 minutes instead of 45.
8 projects. Alf manages them all. ~$3.50/day total. You bill each client for managed services and pocket the difference.
Alf monitors your infrastructure 24/7. Known issues get fixed automatically. Unknown issues get escalated immediately.
Continuous
System Monitor
Health checks across all services. Container status, memory, disk, API latency, error rates. Doesn't just alert — creates actionable tickets with context from past incidents.
Auto-detected
Ops Worker
Redis memory creeping. Checks Alf's memory for what worked before. Finds: "volatile-lru eviction policy resolved this in 4 minutes." Applies it. Zero downtime. No human involved.
Escalation
System Monitor
Unknown error pattern in the auth service. No match in memory. Escalates to Slack and PagerDuty with full diagnostic context: logs, metrics, recent events, and suggested investigation paths.
Weekly
Brain
Correlates incidents: "Redis memory spikes happen every Tuesday after the batch job runs." Creates a proactive ticket to reschedule it. Pattern detection no dashboard can do.
$0.43/night instead of $4,000/month for an on-call engineer. Known issues auto-remediated. Unknown issues escalated with full context.
MEMORY

Alf never makes
the same mistake twice.

Every task, every outcome, every interaction is stored in a cognitive memory that grows with every execution. The more Alf works, the smarter your team gets.

0 knowledge nodes
They remember what they learned

"Redis connects to the rate limiter, which connects to the auth middleware." The agents map your entire system — services, dependencies, failure points — without being asked.

They remember what worked

"Last time the database locked up, increasing work_mem to 12MB fixed it in 4 minutes." Next time it happens, they'll try that first. Automatically.

They get faster every day

Successful patterns are extracted into reusable procedures. The 10th deploy is faster than the 1st. The 100th security scan knows exactly where to look.

Full transparency

The agents don't always get it right on the first try.

A worker once committed a fix that broke a different feature. A tester caught it in 14 minutes, created a new ticket, and another worker rolled it back. The system self-corrected — but the first fix was wrong.

We're telling you this because every other AI product pretends to be perfect. We'd rather show you how the system recovers from mistakes than pretend it never makes them. The ticket system, the monitoring, the verification — they exist because workers are powerful but not infallible.

The value isn't perfection. It's that problems get found and fixed at 2am instead of discovered at 2pm.

TRY IT

Tell Alf what you need.
Alf figures out the rest.

Alf analyzes your request, creates the right specialist(s), and gets to work. Try anything.

8+∞
Core + Custom Agents
16
Channels
47
Integrations
12
Device Types
70
Built-in Tools
THE DIFFERENCE

Not another chatbot.
A self-managing workforce.

Alf ChatGPT / Claude OpenClaw Zapier / Make
Multi-specialist teams ✓ 109 templates + unlimited custom ✗ Single assistant ~ Single agent ✗ No agents
Persistent memory ✓ 3-tier memory + knowledge graph ~ Per-conversation only ✗ No memory ✗ No memory
Self-hosted ✓ Your hardware, your data ✗ Cloud only ✓ Self-hosted ✗ Cloud only
Autonomous operation ✓ Runs 24/7 unattended ✗ Requires prompting ~ Scheduled only ~ Trigger-based
Cost tracking ✓ Per-agent budgets + audit trail ✗ Monthly bill only ✗ No tracking ~ Task-level billing
Team coordination ✓ Ticket dispatch + escalation ✗ No coordination ✗ No coordination ✗ No coordination
Mission control dashboard ✓ Full team visualization ✗ Chat window ✗ CLI only ~ Workflow editor
Embedded AI terminals ✓ Claude CLI + Codex ✗ ✗ ✗
Open source ✓ MIT license ✗ Proprietary ✓ Open source ✗ Proprietary
MARKETPLACE

109 templates. Community library. One-click install.

Browse 16 categories of worker templates — Personal, Marketing, Support, E-Commerce, Finance, and more. Share skills with the community. Import/export bundles.

Tools
Docker API · Security Scan · Code Analysis · Deploy Ops · Team Intel · Web Search · DB Query · Knowledge Graph
Skills
Incident Responder · PR Reviewer · Deploy Manager · Security Scanner · Data Analyst · Cost Optimizer · API Tester
Community
Publish your own packages · Ratings and reviews · Version tracking · One-click install from dashboard
THE DASHBOARD

Your AI workforce dashboard. Not a chat window.

Everything you need to run, monitor, and manage your AI workforce — in one place.

🛰
Ask Alf + Team View
Streaming AI chat with slash commands. Team visualization. Execution stats. Cost tracking. Real-time briefings.
📺
Live Execution Streaming
Watch workers in real-time via SSE. See outputs, tool calls, and results as they happen. Branch review queue for code changes.
⏰
Visual Schedule Editor
Preset pills (30m to 24h) or custom cron. Human-readable preview. Next run time. Per-worker scheduling.
🧠
Memory — Ask, Teach, Browse
Semantic vector search. Teach Alf facts, preferences, and rules. Knowledge graph. Memory analytics and consolidation.
🧬
5 AI Providers + Ollama Local
Anthropic, OpenAI, Ollama (free local), xAI, DeepSeek. OAuth connect from dashboard. Pull/manage Ollama models in Settings.
🛡
Infrastructure Dashboard
VPN config, SearXNG status, Autoheal monitoring, PostgreSQL/Redis health, Ollama model management — all from Settings.
🛒
Marketplace & Community
109 templates. Community skills library with ratings. Tool bundles. MCP servers. Browse, install, rate, and share.
🔐
Security & Encryption
AES-256-GCM encryption at rest. Guardrails with content filters. Rate limiting. Prompt injection blocking. OAuth auto-refresh.
USING OPENCLAW?

Migrate in 60 seconds. Or keep both.

Upload your openclaw.json in Settings > Migration to import everything. Or connect OpenClaw as a channel frontend — its 19 chat platforms, your AI workforce.

OPENCLAW_GATEWAY_URL=ws://127.0.0.1:18789

Feature comparison: AskAlf vs OpenClaw

BUILT INTO THE DASHBOARD

Full Claude CLI and OpenAI Codex terminals.
Inside your dashboard. Not a wrapper.

Open the Workspace tab. Full Claude Code or Codex session with access to your projects, your workers' context, and Alf's memory. No copy-pasting. No context switching.

CODEXCLAUDE CODE
RUNNINGPID 2909
WORKSPACE ▾CTRL+CRESTART
█▌█▌█▌█▌ Claude Code v2.1.76 █▌█▌█▌█▌ Opus 4.6 (1M context) · Claude Max /workspace ──────────────────────────────────────────────── fix the race condition in payment handler ● Found the issue on line 412 of src/handlers/payment.ts. Two concurrent requests can read order.status before either commits. ⎿ Edit src/handlers/payment.ts - const order = await getOrder(id); + const order = await getOrderForUpdate(id); ⎿ Bash npm test -- payment 42 passed | 0 failed | 1.2s Fixed. SELECT ... FOR UPDATE prevents the concurrent read. ──────────────────────────────────────────────── ❯ ❯❯ bypass permissions on (shift+tab)
CODEXCLAUDE CODE
RUNNINGPID 3041
WORKSPACE ▾CTRL+CRESTART
codex v0.98.0 ╭────────────────────────────────────────────╮ │ │ │ scaffold a REST endpoint for webhooks │ │ │ ╰────────────────────────────────────────────╯ Codex I'll create the route, handler, schema, and tests. create src/routes/webhooks/index.ts create src/routes/webhooks/handler.ts create src/routes/webhooks/schema.ts create src/routes/webhooks/webhooks.test.ts exec npm test -- webhooks ✓ 8 passed | 0 failed | 0.9s Route registered at /api/v1/webhooks codex>
THE ECOSYSTEM

12 device adapters. 47 integrations. 16 channels. 109 templates.

Your agents don't just run in the cloud — they control real machines, talk to your tools, and reach you anywhere. Encrypted credentials. Signature verification. Test-before-deploy.

Compute & IoT (12 Adapters)
💻
CLI Agent
Any machine with Node.js — npm i -g @askalf/agent. Auto-reconnect, heartbeat, task streaming.
🐳
Docker Host
Container management, builds, deploys. AES-256 encrypted credentials.
🔐
SSH Remote · Kubernetes
Remote servers via SSH tunnel. K8s pod orchestration. Encrypted keys at rest.
🌐
Browser Bridge · Desktop Control · VS Code
Web automation, screenshots, DOM access. Mouse/keyboard control. Editor integration.
📱
Android · iOS
ADB bridge, app automation. Simulator/device testing, shortcuts.
💡
Raspberry Pi · Arduino/ESP32 · Home Assistant
GPIO, sensors, edge compute. Microcontroller programming. Smart home automation.
Channels (16)
💬
Slack · Discord · Teams
Bot tokens. Slash commands. Signature verification. Auto-test on connect.
💬
Telegram · WhatsApp
Webhook auto-registration. Bot username detection. Platform verification.
🦞
OpenClaw Bridge
Connect OpenClaw as a frontend. Its 19 chat platforms, your AI workforce.
🔗
REST API · Webhooks
Direct API dispatch. Inbound/outbound webhooks with signature verification.
⚙️
Zapier · n8n · Make
Webhook URLs. API key auth. Bi-directional automation workflows.
📧
Email (SMTP) · SendGrid · Twilio SMS
Transactional email. Bulk campaigns. SMS dispatch and receive.
🎤
Twilio Voice · Zoom
Voice calls, IVR routing. Video meeting bots and transcription.
Integrations (47) — 13 Categories
📦
GitHub · GitLab · Bitbucket
Source control. OAuth. PR automation. Pipeline triggers.
💰
Salesforce · HubSpot · Pipedrive
CRM. Leads, contacts, deals, pipeline management.
🛒
Shopify · Stripe · WooCommerce · Square
E-Commerce. Orders, payments, inventory, subscriptions.
📣
Mailchimp · Google Ads · Meta Ads · SendGrid
Marketing. Campaigns, audiences, email, ad management.
📱
Twitter/X · Instagram · LinkedIn · Buffer
Social. Posts, mentions, analytics, scheduling.
📝
Google Workspace · Microsoft 365 · Airtable · Sheets
Productivity. Docs, email, calendars, spreadsheets.
☁️
AWS · GCP · Azure · DigitalOcean
Cloud. Infrastructure, compute, storage, functions.
🚀
Vercel · Netlify · Railway · Fly.io
CI/CD. Deploy, builds, serverless, edge.
📋
Jira · Linear · Notion · Asana
PM. Issues, sprints, boards, knowledge bases.
📊
Datadog · Sentry · PagerDuty · Grafana
Monitoring. Metrics, alerts, incidents, dashboards.
📈
Google Analytics · Mixpanel · Plausible
Analytics. Traffic, funnels, events, conversions.
💵
QuickBooks · Xero · Gusto · Wise
Finance. Invoices, payroll, accounting, transfers.
🗄️
Cloudflare · S3 · Supabase
Storage. CDN, object storage, database, edge.

Runs on Linux, macOS, and Windows. If it has Docker, it runs AskAlf. If it has a network connection, your agents can control it.

Optional: Enable Autoheal for automatic container recovery. Route all agent traffic through Proton VPN via Gluetun — your agents' traffic is encrypted and anonymous. One flag in your env file.

TypeScript 5.4
Node.js 22
React 19
Fastify 5
PostgreSQL 17
pgvector 0.8
Redis 8
Docker Compose
Claude CLI
OpenAI Codex
SearXNG
Cloudflare Tunnel
WebSocket
MCP Protocol
PKCE OAuth
Gluetun VPN
Autoheal
Ollama
QUESTIONS YOU'RE THINKING

Straight answers.

Is this vaporware?

No. It's running right now on real infrastructure serving real users. 109 worker templates, 16 categories, community skills library, streaming AI chat — all production code. Open source, MIT licensed.

How much does it cost to run?

Infrastructure: free on hardware you own. Or ~$5/month on a VPS. AI costs depend on usage — simple tasks cost $0.01-0.05, complex ones $0.10-0.50. Default budget: $15/day with guardrails that prevent runaway spending.

Do I need to set up workers manually?

No. Tell Alf what you need and the right specialist is created automatically. Or browse 109 pre-built templates. Or create custom workers with the 3-step builder. Your choice.

What AI providers does it support?

Anthropic (Claude Opus, Sonnet, Haiku), OpenAI (GPT-4o, o3, Codex), Ollama (DeepSeek R1, Llama 3.2, Qwen 3, Gemma 3, Phi 4, Mistral — fully local at $0), xAI (Grok), DeepSeek. Workers auto-select the best model for each task. Pull and manage Ollama models from the dashboard. Your API keys encrypted at rest.

Can workers take real actions?

Yes — that's the point. They browse the web, write files, run commands, create tickets, send emails, and more. You set the autonomy level: full auto, approval required, or manual dispatch only.

What if something goes wrong?

Monitors detect issues automatically. The ticket system creates a paper trail. Cost guardrails cap spending. Approval gates catch risky actions. You get continuous briefings of everything happening across your team.

Is my data safe?

Your data never leaves your infrastructure. No telemetry, no analytics, no phone-home. Everything runs in Docker on your hardware. Optional VPN routes all outbound traffic through an encrypted tunnel. Open source — audit every line.

How is this different from ChatGPT/Claude?

Those are one mind answering one question. This is an entire team that coordinates, delegates, shares memory, learns from outcomes, and operates 24/7 without prompting. For any industry — not just dev.

UPDATES

What's New

v2.6.0 — Multi-Tenant Workspaces, Security Hardening March 27, 2026

Alf Personal / Alf Business — Isolated Workspaces from One Install

Create multiple workspaces with isolated agents, memory, budgets, and executions. Switch between personal and business contexts in one click.

  • Multi-Tenant Workspaces — Create, switch, and delete workspaces. Each has its own agents, memory, tickets, and budget limits
  • Per-Tenant Budgets — Set daily/monthly spend limits per workspace. Dispatcher enforces automatically
  • Email Reports — Scheduled reports now deliver via SMTP in addition to Discord webhooks
  • Security — Fixed 5 CodeQL alerts: rate limiting, format strings. 0 Dependabot vulnerabilities
  • Agent CLI v2.6.0 — Windows service runs completely hidden via PowerShell
v2.5.0 — Remote Execution, Smart Routing, Shell Fallback March 26, 2026

Remote Device Execution, Smart Mode Routing, Scheduled Reports

Execute tasks on remote machines via WebSocket bridge. Smart routing sends shell commands for $0 and complex tasks to Claude. Daily/weekly fleet reports to Discord.

  • Remote Device Execution — askalf-agent connect <key> --install — connect any machine, OAuth syncs automatically, tasks execute remotely
  • Smart Execution Routing — Auto-detects shell commands ($0) vs AI tasks (Claude CLI). 60+ command patterns recognized
  • Shell Fallback — Devices without Claude CLI run commands via PowerShell (Windows) or bash (Linux/macOS)
  • Scheduled Reports — Daily/weekly fleet summaries sent to Discord webhook or email. Configure in Settings > Reports
  • Onboarding Wizard — Ollama auto-detection, use-case team provisioning (9 categories auto-create the right agents)
  • Save as Template — One-click save any customized worker back to the template library
  • Schema Cleanup — Fixed api_keys view, forge_preferences table, usage_count tracking, updatable views via triggers
  • Capabilities Scan — Server requests system info from agents: CPU, RAM, 18 tools, Claude CLI detection
v2.3.0 — Device Bridge & Daily Releases March 25, 2026

Device Bridge, Budget Enforcement, Marketplace Overhaul

32 commits. Connect external devices, enforce spending limits, create your own tools — and daily releases from here on out.

  • Device Bridge — Agent CLI connects via WebSocket, 14 tools auto-detected per device
  • Budget Enforcement — Global daily/monthly limits enforced, per-agent auto-pause
  • Marketplace — Create, import, export tool bundles, MCP servers, and templates
  • Memory Prune — Delete bad facts directly from search results
  • Ollama Models — Quick-pull DeepSeek R1, Qwen 3, Gemma 3, Phi 4, Llama 3.2
  • Daily Releases — Active development, shipping every day
v2.2.0 — Security & Memory March 24, 2026

Security Hardening + Semantic Memory Search

Full security audit, Docker hardening, and a real semantic search engine for Alf's memory.

  • Brain Tab — Semantic vector search across 400+ memories via pgvector
  • Teach Alf — Store facts, experiences, and rules directly from the dashboard
  • Docker Hardening — CPU limits, read-only mounts, credential encryption, execFileSync
  • 155 Search Engines — VPN-routed via Gluetun with optimized engine weights
  • 8 CodeQL Fixes — Command injection, clear-text logging, rate limiting all resolved
  • 70 Tools — 44 forge + 26 MCP tools, unified versioning across all packages
  • Audit Tab — Fixed and wired to forge, 20+ audit entries flowing
  • CI Passing — All Docker images build clean on every push
v2.1.0 — Launch March 23, 2026

Universal AI Workforce Platform

The biggest release yet. AskAlf is no longer dev-focused — it's a universal workforce for any industry.

  • 5 AI Providers — Anthropic, OpenAI, Ollama (local/$0), xAI, DeepSeek
  • 47 Integrations — CRM, e-commerce, marketing, social, productivity, analytics, finance
  • Real-time Streaming — Watch workers execute live via SSE
  • Visual Schedule Editor — Preset pills or custom cron with preview
  • Community Skills Library — Browse, rate, install community templates
  • Central Marketplace — Submit skills with AI security review
  • Infrastructure Dashboard — VPN, SearXNG, Ollama model management
  • AES-256-GCM Encryption — All credentials encrypted at rest
  • 8 Autonomous Workers — Watchdog, Security, GitHub Manager, Discord Manager, Social Media, and more
v2.0.0 March 22, 2026

Universal Pivot

Transformed from a dev-focused 7-agent fleet to a universal AI workforce platform. 109 templates across 16 categories. Dynamic worker creation for any industry.

v1.0.0 March 17, 2026

Initial Public Release

First public release. 7 core agents, 16 channels, one-line installer, PostgreSQL + pgvector memory system, Docker Compose deployment.

⭐
Star on GitHub
Open source · MIT License
💬
Join Discord
Community support
🛒
Marketplace
Community worker templates
𝕏
Follow @ask_alf
Updates on X / Twitter
Share on X
SELF-HOSTED

Your data. Your agents.
Your infrastructure.

No cloud dependency. No API proxies. No vendor lock-in. Deploy on your own hardware with Docker Compose. Guided 5-step onboarding. Optional VPN tunneling. Automatic container recovery. OAuth built in.

$ terminal
curl -fsSL https://get.askalf.org | bash
Release Notes (v2.5.0) Join Discord View Source