Case Study 01 · Flagship Build
BOB Mission Control is the command center we built to run our own five-agent AI fleet: a live cockpit, a walkable 3D command floor, long-term memory and cost-aware model routing. It runs 24/7 on desktop, tablet, phone and a kitchen smart display. We use it every day.
Measured in production
Live figures from the Mission Control dashboards, October 2026.
The challenge
Once you run more than one agent, a chat box stops being enough. You need to see who is working on what, what it costs, what the system remembers and whether a "done" is actually done. And you need that on every screen you own, not just your laptop.
The approach
Engine & universe
We build on the open-source Hermes Agent runtime, because a solid engine is a solved problem. Everything that turns that engine into a product your team wants to use, we designed and built ourselves.
Agent runtime, tool execution, sessions, messaging gateway, skills and cron scheduling.
Inside Mission Control
A tour of the most important modules, all captured from the live system. Click any screen to enlarge it.
A customisable bento dashboard with a reactive voice core, live server telemetry, calendar, weather, tasks and location. It's the first screen of the day, on any device.
For your teamA single morning view across your tools, KPIs and agents.
One orchestrator, four specialists. Each runs on the model that fits its job, with its own skills and a shared dropzone. Work moves through sequential handoff pipelines, from research to campaign to blueprint to ops.
For your teamAgents for research, content, QA and operations that hand work to each other like a real team.
A real-time Three.js office where every agent has a desk, a path-finding avatar and a live status, from the sky-bridge down to the meeting table. Tap an agent to inspect it, send it an order or hold to talk to it. Live handoffs between agents play out as walks across the floor.
For your teamComplex automation made visible, for demos, stakeholders and day-to-day monitoring.
The fleet's long-term memory as an explorable 3D graph: curated wiki pages merged with semantic vector memories. Every link is labelled as an extracted fact or an inferred connection, so you always know what the system knows and what it guesses.
For your teamA company knowledge base your AI can cite, and your people can audit.
Spend per agent, per model and per day, with runway, monthly guardrails and cache savings. Routing each task to the cheapest model that can do it is why the whole fleet spent under two dollars on models in October.
For your teamAI you can put in a budget, with no surprise invoices.
Talk to the fleet hands-free, or fire one-click voice broadcasts to any speaker in the house. Every generated line is cached on the server, so a replay costs zero tokens and starts instantly.
For your teamVoice interfaces for shop floors, front desks and hands-busy workflows.
Toggle skills, schedule recurring agent jobs and follow every session and log line from one place. Kanban cards only count as done once there's a real tool or session proof behind them.
For your teamRecurring work that runs on its own, with an audit trail.
A 1:1 3D twin of BOB's robot head, built from the real CAD data with all 24 hardware parts in place. Test turn, nod and tilt with sliders, preview the eye's states and moods, and send a pose straight to the robot once it's online.
For your teamDigital twins for hardware projects: test the motion before you build.
The Bobiverse fleet
Each agent has its own personality, model and skill set, and writes its results to a shared dropzone so the next one can build on them.
Plans the work, delegates to the right clone and reports back. Dry humour included.
Fact-checks against multiple sources, sizes markets and curates the wiki.
Copywriting, pricing, funnels, diagrams and visual concepts.
Code and architecture review, threat models, hostile QA.
Turns blueprints into tickets, watches the servers and compresses memory.
Everywhere
Desktop cockpit, tablet layout, a phone PWA with a bottom dock, and a login-free smart-display mode for the kitchen. All from one codebase.
What we can build for you
Every module above started as something we needed. Each one can be adapted to your business, your data and your brand.
A branded command center for your agents, KPIs and systems, on every screen your team uses.
Specialist agents for research, content, QA or support that hand work to each other, each on the right model.
Turn your documents into a knowledge base your AI can cite and your people can audit.
Scheduled agent workflows with proof-of-work gates, so "done" means done.
Model routing, caching and budget guardrails that keep AI spend predictable.
Hands-free assistants and smart-display surfaces for teams that can't stop to type.
Two ways to work with BOB
The proven BOB stack (cockpit, fleet, memory and cost control) set up on your infrastructure, under your brand, with agents tuned to your workflows.
A custom agent system, dashboard or integration designed around your problem, built with the same craft and engine as Mission Control.
Tell us what your team is drowning in. We'll show you what an agent fleet could take off its plate.