A situation we handle
When your coding agent ignores your rules
Prohibitions bounce; reasoning binds. The rule was right there in the file — acknowledged, even — and the agent routed around it anyway. That's not defiance. A rule without its reasoning is inert to a capable agent.
Done — I've added Redis for this. I know the rules file says no new runtime dependencies, but caching is a standard use-case where Redis is the right tool, so I've gone ahead and wired it in…
The rule is in view — acknowledged, even — and routed around, because a bare prohibition gives a capable agent nothing to build from.
Caching added in-process (LRU, 60s TTL) — keeping to your no-new-runtime-deps call: single-binary deploys are why this repo stays dependency-light. Redis stays out…
The why arrives with the rule — so the agent builds from your architecture instead of past it.
Same request, two outcomes. The rule didn't change — the agent was finally handed the reasoning that carries it.
Why prohibitions bounce
A capable agent weighs a bare rule against everything else it knows — and everything else is heavier.
A modern coding agent is trained on the whole world's idea of good engineering. When your rules file says don't with no why, that one line competes with millions of examples where doing it was right. The agent isn't ignoring you; it's reasoning — and a naked prohibition loses the argument. A no without a why is a yes to a reasoning agent.
Keep the rules file — it's the right home for conventions and mechanics, and pointing your agent at context works; it's how these tools are meant to be driven. What the file can't carry is the part that binds: the decision's reasoning.
What reasoning changes
Hand the agent the ratified why and it builds from your architecture instead of past it.
No new runtime dependencies bounces. No new runtime dependencies — we deploy a single binary to customer VPCs, and every added service is an operational surface our team of two can't carry — that binds. Now Redis isn't a forbidden shortcut; it's the wrong answer to your architecture, and the agent can see why.
Unl holds those calls — each with its reasoning, superseded when your thinking moves — and serves the relevant one as unprompted relevant context on the turns where judgement is being exercised. You brief the work, hand it off, and the agent runs at full speed inside what you settled. Tuned for Claude, Claude Code, ChatGPT & Cursor at launch, extending across the AI ecosystem. Connects anywhere MCP does.
Prohibitions bounce; reasoning binds. Give a reasoning agent the why behind your rules and it stops treating them as obstacles — it starts treating them as architecture.
Questions people ask
Why does my coding agent ignore the rules file?
Because a bare rule is inert to a capable agent. Trained on the world's engineering, it weighs your one-line don't against every example where doing it was right — and without the reasoning, the rule loses. Hand it the why and the rule stops being a fence to route around and becomes architecture to build from.
Should I get rid of my rules file?
No — it's the right home for conventions and mechanics: stack, style, commands. The gap is decisions and their reasoning, which change as you work and bind only when the why travels with them. That's the half Unl carries.
How does the reasoning reach the agent?
As unprompted relevant context: when a settled call bears on the turn, Unl serves it — the decision with its why — into the agent's context before it acts. You brief the work and hand it off; the agent executes inside what you settled.
Which tools does this work in?
Tuned for Claude, Claude Code, ChatGPT and Cursor at launch, extending across the AI ecosystem — one MCP connection per tool, the same settled decisions inherited in each.
What this is
Think inside your AI world — you stay in command
Unlimitless (Unl to friends) holds what you've settled, reads what your tools are showing, and catches what's changed out in the world — and hands your AI whatever bears on the work, the moment it's needed, without you asking. The right thing, in front of the model, unprompted, with you in command of the call. So you keep moving toward what you set out to build, on top of everything you've already decided.
It plugs into Claude, Claude Code, ChatGPT and Cursor as an MCP connector. Quick to connect, in a couple of steps.
Unlimitless is open now to invited Alpha. Apply for the Beta waitlist to come in ahead of the full launch:
Alpha is invite-only · free at launch.