Behind the Scenes of Daruma
Learn how Auron Labs assisted Daruma with rapid prototyping, media generation machine, and social integration of an entire multi-agent system design.
How Auron Labs Built Daruma’s AI Engine in Weeks
A few weeks ago, the Daruma team hit us up with something rare in this space: a massive creative vision and a clear idea of where they wanted to land. They didn’t just want some AI tools duct-taped together: they wanted a full-stack, automated, multi-agent system to generate, validate, and publish everything: characters, videos, posts, audio, lore, you name it.
The timing was tight - launching before specific market windows, but the direction was sharp. And honestly, when a team shows up that clear-headed, it makes our job easier. As we like to say:
“A question well defined is already half solved."
The System Blueprint (Pseudocode Edition)
We always start with flow before frontend. Here’s the high-level structure of the system we designed. This pseudocode version helps align both technical and creative minds around what’s happening under the hood.
Visual Overview

The Stack: Tech + Newbie Breakdown
Story Garden (Airtable MCP)
Tech: GPT fills Airtable with characters, arcs, and metadata
Plain: It’s the brain dump and filing cabinet of the entire creative universe.
Trend Scouts (Proprietary AI)
Tech: Scrapes X/IG trends, community signals, and analytics
Plain: Watches what’s blowing up so Daruma can ride the wave.
Creative Swarm (GPT via n8n)
Tech: Modular GPT agents create story, lore, dialogue
Plain: AI writers doing the heavy lifting, like a 24/7 writing room.
Art Production (MJ, Firefly, Runway, Kling, 11Labs)
Tech: Images via Midjourney + Firefly → Videos via Runway/Kling → Voice via ElevenLabs
Plain: Characters and scenes come to life, fully AI-generated and synced.
Claude Validation
Tech: Runs style/tone checks against custom rules
Plain: Think of it as the editor who never sleeps and always sticks to the vibe.
Launch Swarm (CapCut + Telegram)
Tech: CapCut formats content → Telegram auto-posts to X/IG
Plain: Posts go out while you sleep, already optimized.
Monitoring & Feedback
Tech: Engagement data + EQ signals loop back into trend engine
Plain: System sees what hits and adjusts the future content.
Evolution Engine
Tech: Logs hits, mutates styles, prunes flops
Plain: Content that works gets smarter. Content that flops disappears.
The Result
Because of how tight Daruma’s direction was, and how modular our approach is, we were able to go from zero to system in under 3 weeks. The whole machine runs on top of n8n, with Claude and GPT acting as core agents and guardrails.
They’re now pushing AI-powered stories and assets in real time with near-zero human effort. It's scalable, fast, and completely custom to their tone and goals.
Want Something Like This?
Right now we’re not taking new clients, but we’ve opened a waitlist.
If you’re building something ambitious and want Auron Labs in your corner,
submit your email to our waitlist to be notified when slots open up.