HOW IT WORKS

Buy. Brief. Run.
Done.

Four steps to get Tiro and Dojo working for you. The agents come pre-built with frameworks, persona, and memory. You bring the context. They handle the work — picking the right brain for each task automatically.

THE WORKFLOW

Four steps.
No subscription.

Set-up takes about thirty minutes. After that, you talk to the agents and they handle the rest.

01
INSTALL
Download the desktop app
Mac or Windows. One-time purchase, no licence server, no phone-home. The app lives on your machine.
02
BRIEF
Fill your context packs
Life Context Pack for Tiro. Briefing Package for Dojo. Twelve sheets each. About twenty minutes per pack.
03
CHOOSE A BRAIN
Local or BYOK
Run a local model on your machine for free. Or plug in your own API key. Switch any time, per task.
04
RUN
Talk to them. They work.
Drafts go to your approval queue. You sign off. The agents execute — calendar, email, files, all from your machine.
CONTEXT ENGINEERING

The hard work is already done.

Most AI tools hand you an empty chat box and expect you to engineer the prompt every time. Attyn ships the agents already context-engineered — twenty-five-plus frameworks, persona scaffolding, memory architecture, and approval logic — built in. You don't build the agent. You hand it your context, and it adapts to your workflow.

SHIPS IN THE BOX
What we engineered
  • Agent persona & voice (Tiro / Dojo)
  • 25+ thinking frameworks (CPE, NVC, Attachment, DISC, Eisenhower, OKR, RACI, Minto, MECE, Porter, PESTLE)
  • Memory architecture & long-term recall
  • Approval queue & default-deny logic
  • Task classifier & model router
  • Email, calendar, file system adapters
  • Audit log for every action
YOU PLUG IN
What you bring
  • Your business overview & sector context
  • Your team, suppliers, partners, customers
  • Your tone of voice & communication rules
  • Your finances, goals, and KPIs
  • Your relationships, habits, calendar rhythm
  • Your tools (Gmail, Slack, Notion, etc.)
  • Your boundaries — what they can act on alone
Change the workflow. The agents adapt. No re-prompting. No re-training. No re-buying.
SMART MODEL ROUTING

The agent picks the right brain for the job.

Not every task needs a frontier model. Attyn classifies each task by complexity and routes it to the right tier — frontier, capable, or local. Some tasks need speed and breadth. Some need depth and precision. The agent decides which.

"Not every task needs to finish in two minutes. We expect the same of humans — some work takes its own time to finish perfectly. Match the model to the task, not the task to the model."
— THE ROUTING PRINCIPLE
TIER 1 · FRONTIER LLMS
The Tier 1 graduates
Fast. Broad. Expensive. The elite-college minds — used when speed and breadth both matter.
Claude Opus 4
GPT-5
Gemini Ultra
DeepSeek R1 671B
Strategic memos
Multi-step reasoning
Sensitive writing
Anything client-facing
COST / TASK
High
SPEED
Fast · 3–15s
TIER 2 · MID MODELS
The Tier 2 graduates
Sharp. Lean. Cost-aware. Medium-pace generalists who handle most everyday work.
Claude Haiku 4.5
GPT-4o-mini
DeepSeek R1 32B (local)
Qwen 2.5 32B (local)
Email drafts
Meeting notes
Summaries
Scheduling logic
COST / TASK
Low
SPEED
Medium · 5–20s
~70% OF TASKS
TIER 3 · SPECIALISTS
The specialists
Local. Purpose-trained. Takes the time it needs. Finishes the job perfectly because that's what it was built for.
Gemma 3 4B
Llama 3.1 8B
DeepSeek R1 14B
Phi 4 mini
Logging & classification
Memory updates
Local file ops
Sensitive data tasks
COST / TASK
$0
SPEED
Patient · perfect
THE EXECUTION LOOP

You stay in the loop. The agents stay in line.

Every action with consequence goes through the approval queue. You can mark a category as "approve and learn" — the agent watches your decisions and earns autonomy over time.

01
You speak
Type, paste, or attach. Plain language.
02
Agent classifies
Routes the task by complexity and risk.
03
Brain runs
Local or BYOK — the right tier handles it.
04
Plan returns
Draft plus rationale. No silent execution.
05
You approve
Approve, edit, reject. Or auto-approve a category.
06
Agent executes
Sends, schedules, files. Logs it. Done.
READY TO RUN

Stop renting AI tools.
Start running the system you own.