A downloadable ZIP containing a complete agent system: runtime engine (JavaScript), your customized workspace (SOUL.md, agent configs, skills, guardrails), shell scripts to start/stop/monitor, and macOS LaunchAgent files for 24/7 operation. Everything you need to run autonomous agents on your own machine.
You need a machine that runs 24/7 — a Mac mini, a Linux server, or a VPS. The agents run as background services. A Mac mini (from ~€600) or a Hetzner VPS (from ~€5/month) works perfectly.
Your agents use models from OpenRouter — a single API key gives access to GPT-4, Claude, Gemini, DeepSeek, Kimi, and more. The system includes 5-tier smart routing that uses cheap models for simple tasks and powerful models only when needed. Average cost: ~€8/day for the full 8-agent system.
No. These are persistent, autonomous agents that run 24/7 without your input. They have memory (they learn across sessions), tools (they can read files, search the web, call APIs), guardrails (you define what they can and can't do), and they coordinate with each other. They're closer to employees than chatbots.
ChatGPT and Claude are single-session tools — they forget everything between conversations. Our agents have persistent memory, scheduled execution cycles, tool access, inter-agent coordination, cost optimization, and safety guardrails. You configure once, and they work autonomously from then on.
Two costs: (1) AI model usage — paid directly to OpenRouter, typically €3-15/day depending on your budget tier. (2) Optional updates — your purchase includes 2 months of free updates; after that, 9.99€/month for new features and skills.
Yes — the agents are defined in Markdown files (SOUL.md, AGENT.md). You can edit their identity, add skills, modify guardrails, and change model assignments. The workspace is yours to customize.
Yes. Every 5 completed tasks, the reflection engine reviews what happened — extracting facts learned, mistakes made, and improvements for next time. These are written to persistent memory AND stored as semantic vectors in Qdrant. When facing a new task, agents search their entire knowledge base by meaning — not just keywords. Even better: when an agent repeatedly struggles in a domain with no matching skill, it proposes creating one. You approve, and the agent permanently gains that expertise. Your system literally upgrades itself.
27 domain skills ship with every system: SEO audit, copywriting, content strategy, cold email, analytics tracking, schema markup, page CRO, form CRO, sales enablement, revenue operations, client operations, and more. Skills are auto-matched to tasks by keywords — no manual loading. Agents can also propose new skills when they detect knowledge gaps.
All major platforms: macOS (LaunchAgents), Linux (systemd), Docker (any OS), and Windows (batch scripts). Every download includes a docker-compose.yml with Qdrant vector memory + agent runners — one command to start everything. You can also run with the Klawty Gateway for 21+ messaging channel support.
Docker is recommended for running Qdrant (the vector memory database), but not required. Without Docker, your agents use file-based memory (MEMORY.md + session logs) which works perfectly well. With Docker, you get semantic search across all agent knowledge — agents can find relevant past learnings by meaning, not just keywords. Setup is one command: docker compose up -d.
No. Everything runs on your machine. Your data, your API keys, your conversations — all local. We never see your agents' output or your business data.
Yes. For businesses that need a pre-fitted industry system (restaurants, real estate, construction, etc.), we offer managed hosting from 199€/month — we deploy, maintain, and monitor your agents for you. For custom integrations (ERP, CRM, proprietary APIs), our engineering team at dcode technologies builds and maintains the system. See our industry solutions at /solutions or contact us at d-code.lu.
Every download includes a QUICKSTART.md guide with step-by-step instructions. Setup takes about 10 minutes. For additional support, enterprise customers get dedicated engineering assistance.
Yes — within 14 days of purchase if you haven't been able to get the system running. Contact us and we'll help troubleshoot or process a refund.
Still have questions?