Comparison
Unl vs the CLAUDE.md method — an honest comparison
If your project's thread keeps going slack between sessions, the usual fix is a hand-written CLAUDE.md. It's a decent tool — here's where it holds up, where it doesn't, and how it compares to Unl.
Written to be fair: CLAUDE.md gets its due, and the cons of Unl are on the page too.
| The CLAUDE.md methoda file you maintain | Unla layer that inherits | |
|---|---|---|
| Setup | Write a markdown file by hand | ✓Add the connector — about five minutes |
| Cost | ✓Free, built in | ✓Free at launch; paid tiers later for heavy use |
| Conventions and mechanics (stack, style, commands) | ✓Strong — the right home for these | ~Keep them in the file; Unl doesn't need to own them |
| Current work — where things actually stand | ✗Rots silently between hand-edits | ✓Updates as you settle decisions |
| Holds the why behind a decision | ✗Captures the what, not the why | ✓Every decision with its reasoning |
| Carries what you ruled out (dead-ends) | ✗Not its job | ✓So the agent won't re-propose it |
| Surfaces on point, unprompted | ✗Loaded once at session start, then static | ✓Arrives the moment it bears on the work |
| Maintenance | ✗Ongoing grooming, by hand, forever | ✓None — it holds it for you |
| One source across your AI tools | ~Per-repo file, one tool's convention | ✓One connection over MCP — inherited everywhere you work |
The CLAUDE.md method
Pros
- Free and built into Claude Code
- Zero install; lives in your repo, version-controlled
- Genuinely the right home for conventions, mechanics and house rules
Cons
- You write and groom it by hand — forever
- Anything with a present-tense truth-value rots silently between edits
- Captures the what, not the why behind a decision
- Doesn't hold what you ruled out
- Loaded once at session start, then static — never surfaces on point
Unl
Pros
- No file to write or maintain — holds decisions as you settle them
- Every decision with its reasoning
- Surfaces the relevant one unprompted, the moment it bears
- Carries the dead-ends so the agent won't re-propose them
- One connection over MCP — inherited across your tools
Cons
- It's a connector you add — one more thing in your stack
- Newest in the category; earliest days
- Free at launch, with paid tiers later for heavy use
A file is the right home for conventions and mechanics — keep your CLAUDE.md for those. It's the wrong home for anything with a present-tense truth-value: current work, settled decisions, what you ruled out. That content rots silently between hand-edits, because it's a file you maintain, not a layer that inherits. Unl holds the decisions with their reasoning and surfaces the relevant one unprompted when it bears. That's the gap Unl fills.
The seam, plainly
A file carries the mechanics. A layer carries the truths.
CLAUDE.md earns its place: stack, style, commands, house rules — stable facts that change when the repo changes. The trouble starts when the file is asked to carry things that are true now — where the work stands, what was settled last week, what was tried and ruled out. Those change between every session, nobody re-grooms the file in time, and a stale line reads exactly like a current one.
Unl takes that half of the job: decisions held with their reasoning, updated as you settle them, and handed into the agent's reasoning before it acts — unprompted. You author the brief, hand it off, and the agent executes inside what you settled — full speed on the mechanical work, and done stays yours to say.
Tuned for Claude, Claude Code, ChatGPT & Cursor at launch, extending across the AI ecosystem. Connects anywhere MCP does.
Questions people ask
Is Unl a replacement for CLAUDE.md?
Not exactly — they do different jobs, and they compose. CLAUDE.md is the right home for conventions and mechanics, loaded at the start of a session. Unl is a layer that holds the decisions you've settled, each with its reasoning, and surfaces the relevant one unprompted the moment it bears on the work. Keep the file for the stable facts; use Unl for the decisions and the why behind them.
What does the CLAUDE.md method do well?
CLAUDE.md is free, built into Claude Code, needs no install, and lives in your repo under version control. It's genuinely good for stable context: your stack, your coding conventions, your house rules. Where it struggles is anything with a present-tense truth-value — what you decided and why, what you ruled out, where the work stands — and surfacing any of it at the moment it's relevant.
Why does a CLAUDE.md go stale?
Because you maintain it by hand. The file doesn't update itself, so anything in it that's true now — current work, latest decisions — is out of date by your third session unless you stop and groom it. It also captures the what, not the why; the reasoning that stops a settled question being reopened isn't there to inherit.
Is Unl harder to set up than CLAUDE.md?
It's a different kind of effort. CLAUDE.md is free but you write and groom it yourself, forever. Unl is a connector you add once — about five minutes — and after that it holds your decisions and surfaces them for you, with no file to maintain. Unl is free at launch.
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.