Context and memory
Who you are, how you work, which projects matter, and what the agent should not need explained again.
You answer a questionnaire about your Mac, your tools, your model access, your privacy boundaries, your ethical preferences, and your workflows. Aksel generates 13 setup files specific to your setup — ready to use, hand to an agent, or implement step by step.
The operating layer
An agentic workflow becomes personal when the agent starts from the same operating rules every time: your context, limits, approvals, privacy routing and preferred way of working. The LLM does the reasoning. The harness connects files and tools. The Aksel file pack tells both how to behave.
Who you are, how you work, which projects matter, and what the agent should not need explained again.
What may be touched, what stays local, when approval is required, and which ethical boundaries should travel with the task.
Repeatable steps for useful work, with logging and review so automation stays visible before it becomes trusted.
Aksel is a blueprint product, not a managed service. This boundary is the product.
You implement the setup yourself, or hand the files to a coding agent. Aksel provides the map — you decide what runs.
The blueprint pack
Every file is generated from your questionnaire answers — not a generic template. Simple local files, no proprietary platform, no expiry date.
The questionnaire is where you define what may run locally, what may use cloud AI, which models or subscriptions you have, what the agent must ask before doing, and which ethical boundaries should be written into the setup. If you are unsure, choose Make it work for me and Aksel will start from conservative defaults you can review.
These files are the starting point for workflow automation, not just documentation. They give you the routing rules, playbooks, and agent context to make repeatable multi-step work delegable and auditable from day one.
The master document. One file that describes your complete local-first AI setup — models, tools, routing rules, and conventions.
Drop these in any project root. Claude Code, Codex, and Cursor read them automatically — they know your conventions before writing a line.
Explicit rules: what runs locally, what can go to cloud AI, and what must never touch AI at all. Based on your data types and risk tolerance.
Step-by-step playbooks for your most common AI-assisted work loops — for you to follow, or for an agent to execute.
Ordered install steps for your specific tools and models. No guesswork, no YouTube tutorials — just the sequence that fits your machine.
RAM, unified memory, disk, and thermal assessment. Confirms which local models your Mac can run at usable speed.
A concrete before/after baseline for the workflows you identified — so you can measure whether your AI setup is actually saving time.
A lightweight upkeep checklist: model updates, disk hygiene, context file reviews. Keeps your setup from decaying over time.
Documents what each tool in your setup is trusted to access — files, APIs, clipboard, network. A record you control.
A plain-text diagram of how data moves through your setup — which tools see what, and where the privacy boundary sits.
A local durable log for every agentic project. Records date, actor, goal/prompt, files touched, commands run, model and tool used, decision made, outcome, evidence, risk flagged, and follow-up date. Yours to fill — not monitored by Aksel.
Why the event log matters
When an AI agent works without you watching every step — completing tasks, editing files or using connected services — the event log becomes your record of what happened: what ran, who authorised it, what changed, and what should be checked.
After purchase
The blueprint pack is designed to be used in whichever mode fits how you work.
Drop CLAUDE.md, CODEX.md, or AGENTS.md in your project root. Drop install_plan.md into Claude Code or Codex and ask it to implement. The files are written for this.
Work through install_plan.md and workflow_playbooks.md step by step. Each step is explicit — no assumed context from a tutorial that may be outdated.
Keep the files in a folder on your Mac. Edit them as your setup evolves. They are simple local files — no app required, no sync dependency.
The data_flow.md and trust_manifest.md are written to be readable by a third party — useful if you need approval from IT or a compliance team.
Important — read before you buy
You are buying the instruction file pack, not a guaranteed working system.
Aksel delivers blueprint and setup files. You are responsible for reading and executing any instructions, running any agent actions, granting permissions, connecting tools, and handling all data on your own machine. Aksel has no presence on your system and cannot control what happens when you or an agent acts on the files.
Be aware of the following risks before you proceed: