*Building in the open*
Most companies keep their marketing and technical plan internal. We're publishing ours — positioning, growth strategy, and the honest state of the product — along with the rules we hold ourselves to so none of it drifts into overclaiming. This page is kept in sync with the code, not written once and left to go stale.
Where we're headed
Campfyrr is the collaboration tool for vibe coding — the shared page a team plans on together before anyone opens the IDE.
The problem, stated plainly: AI coding agents make individual developers faster and make the team messier — not because anyone is careless, but because the fast conversation with the agent is private. PRs get huge because nobody split the work first. Context is trapped in whoever's chat with their agent. Architecture gets derailed before anyone reviews a line — one developer tells Cursor to refactor auth, another tells Claude Code to add OAuth on the old schema, and both find out too late.
What's true today
Each of these is demonstrable with a real doc, not just described:
- Command-driven, not chat-driven. No sidebar, no toolbar — press Enter on empty space, type an instruction, the agent writes into the doc live.
- Provenance on every block. Hover any block to see who — or which agent, with what prompt — wrote it.
- Zero-auth entry. No signup wall. Paste notes, land in a live doc, share the URL.
- Sandbox + non-destructive history. Fork a doc into an isolated sandbox to try a risky rewrite, promote it back, or scrub through revision history and restore.
- True multiplayer. Colored cursors and a shared agent-cursor badge, visible to every peer identically — not just "AI plus one human."
- Plain Markdown underneath. No proprietary format — Claude, Codex, and any other AI agent read and write the doc cleanly.
- Real sign-in and per-document visibility. Sign in via Clerk for a durable identity, and switch any document you own to private, unlisted, or public — enforced on every read and write, not just a UI toggle. See Privacy & sharing for exactly what each mode does.
What's not built yet
- No per-person sharing or roles. Private means owner-only — there's no "shared with these five people" mode between that and fully public.
- No automated test suite yet. Changes are verified manually.
- Single LLM provider. Agent generation runs on Groq only today — no fallback or multi-provider routing.
- No community space stood up yet. Still a goal, not a shipped thing.
The rules we publish under
The failure mode we're actively guarding against is a marketing page that's ahead of the product. So:
- Pitch what's shipped, not a roadmap. We don't imply automated conflict-detection or dependency-graph validation exists — today the doc surfaces misalignment when a person reads it, not a tool that finds it for them.
- No "production-grade" or access-control claims we can't back up. If we say a document is private, that means enforced on read and write, not just hidden from a listing — and if it isn't enforced, we say so instead.
- This is a living document, not a snapshot. When the code changes what's true, this page gets corrected — not left to quietly go stale.
> This page is generated from the same planning document the team works from internally (docs/marketing/campfyrr-marketing-plan.md in the repo) — publishing it here means the public version and the internal version can't drift apart silently.
How we grow
- Product-led loops. The share-to-signup loop — a non-owner opening a shared doc can start their own copy in one click, no signup — plus the public doc gallery as an entry point.
- Compounding content. Comparison pages, use-case walkthroughs, and technical writeups, each anchored in a real doc — see Guides. Run as a sustained cadence, not a launch burst.
- Community as an asset. One space, pitched around coordinating a team that vibe codes rather than as product support — not started yet.
- Narrative that compounds. The same one-line positioning everywhere, plus an honest quarterly "state of Campfyrr" update on what's fully true, partial, or not built.
| Channel | Compounds? | Notes | | --- | --- | --- | | Duplicate-doc / share loop | Yes — grows with usage | Shipped; the current focus is instrumenting it, not building it | | SEO content (comparison, use-case, technical) | Yes — grows for months per piece | The long-term engine; needs sustained cadence, not a launch burst | | Community | Yes — value grows with size and activity | Not stood up yet | | Public doc gallery | Yes — grows with every addition | Doubles as SEO and onboarding, not a templates product | | Show HN | No — one-shot spike | One clean shot; hasn't happened yet | | X/Twitter thread | Partial — only if it builds a following | Amplification, not a standalone tactic | | Reddit | No — one-shot per post | A feeder, not a weekly channel |
What's next
- ~~Scope and ship the "start your own" duplicate-doc CTA.~~ Done — now a measurement problem: track the share of new docs it produces.
- Keep the content cadence going. The first four guides are published; the engine only works if it doesn't stop there.
- Stand up the community space. Not started.
- ~~Decide the public answer to document privacy.~~ Answered — private/public/unlisted now exists and is enforced. Still open: per-person sharing.
- Run the launch push — Show HN, a Twitter thread, and relevant subreddits — once the community space exists to catch the traffic.
The technical side
The mechanics behind the live, multiplayer, agent-native editing — the command-line trigger, the two sources of truth, sandboxing — are written up separately as a build-in-public technical deep-dive:
Inside Campfyrr: how it actually works
Same rule applies there as everywhere on this page: it lists what isn't built yet — including the single LLM provider and the missing test suite — right alongside what is.