Kettel — v0.1 — AI-native delivery, with the whole team in the room

Great work comes from
strategy, creative,
and technology
making together.

AI-native delivery broke that partnership. Kettel rebuilds it — a shared plan where all three voices think together, and the agent ships the synthesis.

strategy× creative× technology
01 — the problem

A team of three perspectives, flattened into one person's chat window.

Every enterprise trying to scale AI-native delivery hits the same wall. The creative partnership that produces excellent work — the back-and-forth, the challenge, the yes, and — gets flattened into one person's conversation with an LLM. The rest of the team is a spectator. Intent gets lost. Craft gets diluted.

The handoff tax

Someone defines what matters. Someone else interprets it into direction. A third translates both into prompts. By the time code ships, there are four layers of telephone game between what was wanted and what got built.

The unicorn problem

People who span strategy, creative, and technology are rare, burn out fast, and don't scale. Every AI coding tool on the market is built for that unicorn. Kettel is built for the team that replaces them.

02 — the insight

The best work happens when three voices are in the same room, sharpening each other.

strategy
Why does this matter? Who does it serve and what does "good" look like if we ship it?
creative
What's the bold move here? Where's the moment of delight we'd be sad to leave on the floor?the answer shouldn't be a page, it should be a conversation.
technology
What becomes possible with the tools we have this quarter that weren't in the room last quarter?if it's a conversation, we can stream it — no page load, no skeleton states.
The plan file is the room. The code is downstream.
03 — the artifact

A plan file, with three voices attributed and three colors of ink.

A Kettel plan is a structured markdown file in your git repo. Each section has an owner. Each annotation has a perspective. The agent reads all of it — strategy's intent, creative's direction, technology's approach — and produces code that reflects the synthesis, not just the last prompt.

It's a handoff artifact: self-contained enough for Claude Code, Cursor, or Windsurf to read and execute without needing access to Kettel at all.

plans / ingest-pipeline.md◐ creative is editing

Ingestion pipeline — v0.1

strategy — intent

A new user can drop in any file and see it become a linked notebook entry within 10 seconds. No format coaching.

"any file" is the promise. let's make the drop zone the whole screen, not a card.
creative — direction

The conversion is visible — the user watches a PDF become paragraphs, an image become a described frame. Progress is narrative, not a bar.

narrative conversion = a teachable moment. good.
technology — approach

Vercel edge functions per format, streamed to the client. Vision model for images. Markdown write-back to the repo is atomic.

streaming means we need SSE not fetch — noting for implementation.
04 — upstream

The notebook is a graph, not a folder.

Research, customer quotes, references, spikes, debates — atomic entries that link to each other and to the plan. The agent traverses the graph, not a pile.

customer interview
"We lose the why by the third handoff."

R. — director of platform at mid-market SaaS. Signal: intent decay is the primary complaint.

strategy note
Acceptance criteria as handoff

What if the acceptance criteria is the test suite the agent runs?

creative reference
Figma's comment model

Threads anchored to regions, not files. Context travels with the work.

spike
MCP as the agent bridge

If the plan is the contract, MCP is the socket. v0.2 target.

debate
CRDT vs section lock

Section locking wins for MVP — CRDTs revisit when build mode ships.

05 — how it works

Collect. Collaborate. Synthesize. Hand off.

01

Collect

The team gathers knowledge in the notebook. PDFs, URLs, screenshots, Slack threads — auto-converted to linked markdown.

02

Collaborate

Strategy writes intent. Creative writes direction. Technology writes approach. Section locking. Inline annotations. Every voice has attribution.

03

Synthesize

The agent reads the full plan and linked notebook entries. Generates implementation steps that respect constraints from all three perspectives.

04

Hand off

Self-contained plan handed to any coding agent — Claude Code, Cursor, Windsurf. The plan is the interface contract. No lock-in.

06 — market

The window is open — and it won't stay that way.

>$4B
AI coding tools market. 70%+ held by three single-player tools.
75%
of enterprise workspaces already have coding agents installed.
6–12m
before incumbents ship meaningful team collaboration.
0
tools where strategy, creative, and technology co-author a plan.

Category forming. Aviator coined "multiplayer AI coding." YC S26 RFS names the "deciding what to build" gap explicitly.

Spec-driven validated. GitHub Spec Kit, Amazon Kiro, Thoughtworks Radar — the industry is converging on plans-before-code.

Incumbents locked in. GitHub defaults to issues. Cursor to the IDE. Augment to engineers only. None are plan-native.

The shared artifact. The plan is the viral loop. Every person who receives a link becomes a user. Distribution is built into the format.

the analog

Figma wasn't the best design tool. It was the one that made design multiplayer. That was enough.

Illustrator had better brushes. Sketch had better pixel precision. Figma had a shared URL. Designers stopped exporting PNGs to Slack because the artifact lived where the conversation lived. The medium ate the incumbents.

The same shape is forming now. The best single-player AI coding tools already exist. What doesn't exist is the shared canvas — the place where strategy, creative, and technology meet before code is written. Kettel is that canvas.

07 — the signal

Teams are telling the market they need this.

42%
of committed code is now AI-generated. Teams can't review fast enough. The coordination gap widens every quarter.
96%
of developers don't fully trust AI-generated code. Trust gets built through shared intent, not post-hoc review.
Having 8 agents write code all day generates more code in a day than one person can review in a week.
— Hacker News, 2026
An effective spec-driven-development tool would have to provide a very good spec review experience.
— Martin Fowler, 2026
What's missing is how to share all that context. I don't think any one current solution will be the winner.
— HN, on team collaboration with coding agents
The intent behind decisions gets lost. You're left with a codebase nobody fully understands anymore.
— Red Hat Developer, "The Uncomfortable Truth"
08 — business

Open core. Expand by surface.

core — open source

The plan file, free.

Plan file, notebook, section locking, annotations, agent discovery + plan modes, GitHub integration, ingestion pipeline.

v0.1Web application — the room, in the browser
v0.2MCP server — technology accesses Kettel from the IDE
v0.3CLI tool — power users and CI/CD pipelines
v0.4API — third-party integrations and custom workflows
The shared artifact is the viral loop. Every person who receives a plan link becomes a user.
let's build the room

Let's think together.

Craft doesn't scale through faster execution. It scales through better thinking. The more powerful the agent, the more the plan matters — and Kettel owns the plan.