# Unl, your why agent (unlimitless.ai) > The why agent for your AI stack. Unl keeps what you decided and why, what changed and what still holds, then brings the part that bears into Claude, ChatGPT, Cursor or whichever AI you're working with. Memory keeps what. Unl keeps why. ## Why, not rules Your AI already knows how to reason. Unl gives it your reason. Without your why, your AI gives the average answer. With it, it gives yours. Tell an AI "exports run through the queue" and it will obey, then guess when it meets an export nobody mentioned. Tell it why, "sending them directly failed on large accounts", and it can work that case out. A small export carries no risk, so it sends it directly, and a large one goes in the queue. A rule covers the cases you listed. A why covers the ones you didn't. Unl keeps each decision a person makes, with its why and with what it replaced. When their AI starts a task, it gets the few decisions that bear on that task, in the order they were made, never everything they ever said. So the AI stops asking what the person already decided, and comes back to them only when their decisions run out or two of them conflict. One sentence from the person carries into work they never listed, in every AI they use. And a stronger model takes the same why further, so the record is worth more with each new one. ## Why a kept why goes further A why that is not kept when the decision is made is gone, and no model can work it out later, only guess at it. A why that is kept can be walked, because each decision points back to what it came from and to what it replaced. The walk gives direction, so an AI can see where the work is heading and not only where it stands. Direction lets the AI take more steps on its own, which gives it more autonomy, not less. And this grows as models improve. A kept why is not a rule built by hand that a stronger model outgrows. It is material a general model works from, so the stronger the model, the further it goes. Keep the why, and your AI can go further without you. ## Set up from your agent A person can paste this one line into their AI agent, and the agent sets Unl up by following [start.md](https://unlimitless.ai/start.md): Read unlimitless.ai/start.md, set up Unl, and show me the decisions you find. ## Pages - [Home](https://unlimitless.ai/): what Unl is, shown in a fresh session. - [Why](https://unlimitless.ai/why): a no without a why is a yes; how keeping why differs from memory, what it requires, and why it compounds. - [Compare](https://unlimitless.ai/compare): Unl against custom instructions, CLAUDE.md and rules files, memory features, MCP memory servers and running your own, in one table with a section per row and plain answers. - [How it works](https://unlimitless.ai/how-it-works): purpose, the decisions that bear with their why, reported work, inherited outcomes and new human judgement. - [For individuals](https://unlimitless.ai/individuals): decide once in any chat, and every AI you use starts from what you decided, and why. - [For teams](https://unlimitless.ai/teams): the organisation's decisions, each team's why and each person's own work, one why reaching every agent. - [For platforms](https://unlimitless.ai/platforms): bringing each user's decisions, and why they made them, into the AI a platform already ships. - [Unl in Slack](https://unlimitless.ai/slack): when your agents need a decision, answer in Slack, and the why goes with the answer. - [AI-native, at work](https://unlimitless.ai/live): the canonical operating story through real work. /ai-native redirects here. As Markdown: [live.md](https://unlimitless.ai/live.md). - [For builders](https://unlimitless.ai/builders): you've built it, now what. - [Orchestration](https://unlimitless.ai/orchestration): Unl with the orchestrator you already run. Your orchestrator moves the work, Unl carries your why. OpenAI Agents SDK, Vercel AI SDK, CrewAI, Microsoft Agent Framework, Google ADK, LangGraph, Mastra and n8n, each labelled by how far the fit is proven. - [Developers](https://unlimitless.ai/developers): the short page for people: what Unl does for your agents, the first call, MCP and the API. As Markdown: [developers.md](https://unlimitless.ai/developers.md). - [Connect](https://unlimitless.ai/connect): adding Unl to Claude, ChatGPT, Claude Code, Cursor and other MCP clients. - [Pricing](https://unlimitless.ai/pricing): the tier ladder, calibrated to the AI plan you already pay for. - [Trust](https://unlimitless.ai/trust): what Unl can and cannot do, where authority sits, what is kept and for how long, and how to export or delete everything. - [Writing](https://unlimitless.ai/writing): posts on why Unl is built the way it is. - [Never bet against the reasoning](https://unlimitless.ai/writing/never-bet-against-the-reasoning) (19 October 2026): a rule lists cases. A reason generates them. Give a reasoning machine the thing it is built to use. - [A record made of reasons](https://unlimitless.ai/writing/a-record-made-of-reasons) (11 October 2026): without your reasons, your AI reasons toward the average answer. With them, it reasons toward yours. - [The better the model, the more your why matters](https://unlimitless.ai/writing/designed-for-the-frontier) (10 October 2026): every new model can work out more for itself. What you decided, and why, is the one thing none of them can. ## For developers and their agents - [Set up Unl for this project](https://unlimitless.ai/start.md): for an agent the person asked to set up Unl. Add it, hand them the sign-in, read the project's instruction files, and keep only the decisions they say are right. - [The full developer reference](https://unlimitless.ai/llms-full.txt): every endpoint, response, error, scope, MCP tool and worked example, as plain text. - [The live schema](https://api.unlimitless.ai/api/v1/schema): the API's own published JSON Schema, read with your key. - MCP, one address for every client: https://api.unlimitless.ai/mcp Citation welcome.