Collaboration with floww

On this page

A Different Kind of Collaboration

floww is not built for real-time co-editing. It is built for shared understanding — the ability to pick up another person’s work, or return to your own work after a long break, without losing any of the context that makes the work legible.

When Claude is your primary collaborator, the bottleneck is rarely “can we work at the same time?” The bottleneck is “do we understand what was done and why?” floww answers this with three mechanisms: stamps that capture state and intent, branches that preserve alternative paths, and a planning model that makes decisions explicit.

The goal is zero-loss handoffs. When you hand a project to a collaborator — or to a fresh Claude session — nothing should be lost that was not intentionally discarded. Every decision, every explored path, every piece of context that shaped the current state should be recoverable from the artifacts floww produces.

Stamps as Shared History

A stamp (a context-aware snapshot of a workfloww) is not a commit. It captures more than files: it records the canvas state at the moment of creation, the conversation history that led there, and an intelligent summary explaining what changed and why. When you or a collaborator returns to a project, the stamp timeline tells the story of how the work evolved — not just what the codebase looks like now, but the reasoning and exploration that produced it.

Stamps are immutable. Once sealed, they cannot be changed. This makes them reliable anchors: you can always return to exactly that moment, with exactly that context.

The stamp timeline on the canvas is navigable: click any stamp to see its summary, inspect the conversation that led to it, or restore from it. The visual timeline is a shared artifact — everyone who opens the project sees the same history, in the same order, with the same context attached to each point.

Branching as Exploration

When you want to try an approach without committing to it, you branch from a stamp. The branch creates a new workfloww (named sequence of connected nodes on the canvas) that shares the stamp’s history but diverges from that point forward. On the canvas, branches appear as visually distinct paths — you can see both the original direction and the experimental one simultaneously.

If the branch works, it becomes the main line of work. If it does not, you archive it. The original path is always untouched. Branches make it safe to explore: there is no “undo” that overwrites history, no way to accidentally destroy what came before.

The Planning Model as Communication

For larger projects, the planning model (structured hierarchy of milestones, phases, and plans) is also a communication layer. The artifacts it produces — CONTEXT.md (decision record), RESEARCH.md (technical findings), PLAN.md (execution instructions), SUMMARY.md (what was built) — are designed to be readable by anyone joining the project.

A collaborator who arrives mid-phase reads the CONTEXT.md to understand the decisions made and the scope boundaries. They read RESEARCH.md to understand the technical landscape. They read the PLAN.md to know what is being executed. When the plan completes, SUMMARY.md records what was built and any deviations. No verbal briefing needed. No Slack thread to excavate.

This artifact-first approach also applies to Claude. When an executor agent picks up a plan, it reads the same files a human collaborator would: the context for why this work exists, the research that shaped the approach, the plan that specifies what to do. The artifacts serve both audiences — human collaborators coming up to speed and AI agents continuing execution — with exactly the same files.

Why It Works This Way

Real-time co-editing solves the wrong problem for AI-assisted work. When Claude is your collaborator, the bottleneck is not “can we type at the same time?” — it is “do we understand what was done and why?” Stamps solve this: they record not just state but intent. The planning model solves this: it records not just instructions but decisions. Branching solves this: it records not just the chosen path but the alternatives that were considered. floww is designed for asynchronous understanding — the ability to come back later (hours, days, weeks) and pick up the full context of where a project was and why it was there.

The design principle here is durable context. Every piece of information that shaped a decision should be preserved in a form that survives session boundaries, context compaction, and team changes. The artifacts floww produces are not documentation written after the fact — they are produced as part of the work itself, which means they are always current and always complete.

In Practice

Scenario

You work on a feature all morning with Claude, creating 8 stamps across two workflowws. Your collaborator opens the project in the afternoon. They do not need a standup meeting or a Slack summary. They zoom out on the canvas and see the morning’s work: two workflowws branching from the same starting point, stamps marking the progression. They click through the stamp timeline — each stamp has a summary explaining what changed and why. They read the CONTEXT.md from the planning phase to understand the original decisions. In fifteen minutes, they have full context. They create a new workfloww from the latest stamp and continue the work. No information was lost. No context was verbal.

Solo Collaboration

Collaboration in floww is not only about multiple people. The most common collaborator is yourself — your past self who made decisions you do not remember, your future self who will pick up this work in two weeks. Stamps and planning artifacts are as useful for solo developers as they are for teams. When you return to a project after a break, you do not reconstruct context from memory or git log. You read the stamp summaries. You open the last CONTEXT.md. You are current in minutes.

This is why floww treats shared understanding as the primary collaboration problem. Whether the collaborator is a teammate or your future self, the goal is the same: arrive at full context without a verbal briefing.

The tools that serve this goal are stamps, branches, and planning artifacts. They do not replace communication — they make the essential communication durable. The difference between a project that is hard to hand off and one that is easy is not the quality of the standup meeting. It is whether the decisions, explorations, and reasoning are recorded somewhere reviewable.