For engineering leads running a team on Claude Code
By Soumen Sarkar · October 2026
For you. If you lead a team on Claude Code, you are probably the router: sessions don't know about each other, questions wait for you, and pull requests arrive without their reasons. Campfyrr briefs every session, routes assumptions to the person who knows, and puts reasons on every change. A team can set it up in about an hour.
What does a lead's day look like right now?
You own how a team of three to twenty people uses Claude Code. Everyone runs two or three sessions. Each session is fast and capable, and the work it does is good. The trouble sits between the people running them.
A typical day goes like this:
- Sessions don't know about each other. Two sessions change the same part of the code in different directions. Each one loaded CLAUDE.md, and neither saw the other's plan. See why sessions don't know what each other are doing.
- You are the router. A session meets something it can't know from the code and asks. The developer passes the question to you. A question in Claude Code pauses the work until someone answers, and that wait "can stay pending indefinitely" (Agent SDK docs). By lunch you've answered a dozen questions and written no code.
- Review waits on the why. A pull request shows what changed. You still have to work out what it's for, which calls it made and what it took for granted.
None of this is the agents' fault. They are the best collaborators we've ever had. What breaks is the path between people.
What changes with Campfyrr?
Campfyrr is shared team memory for Claude Code. Three things change for you.
Sessions arrive briefed
The first time a session reads or edits a part of the code, it gets an area brief: work in progress there, the decisions that apply, and what was tried and dropped. Work in progress carries its plan, so a second session builds on it instead of around it.
Questions become assumptions that go to the right person
When something is unclear, a session states what it's assuming and keeps going. Campfyrr checks the team's record first. If the record doesn't answer it, the assumption goes to whoever made the closest decision, or to the last person who worked there. Each open assumption has its own page, where people keep it, correct it or pass it on. The answer reaches the session at its next turn. You keep only the calls that are yours.
We run this way ourselves. This repository tells every session: "Turn unknowns into calls: when something is unclear, `stateassumption` with what you're assuming and keep going. Corrections arrive at a later turn."_
Every change carries its reasons
When a session commits, the hook writes a reasons page from what the session recorded: what the change is for, the calls made, the assumptions, what was dropped. The link goes in the pull request description. Read it before the diff. See review is the bottleneck.
What do you look at each day?
Each project has one Activity page. It is a read-only feed, so there are no buttons to work through. Filter it by type, person, area, or "since your last visit".
Spend ten minutes on it each morning:
- Open Activity and filter to "since your last visit".
- Read the "What Campfyrr caught this week" card.
- Open each assumption still open in your areas. Keep it, correct it or pass it on from its page.
- Check decisions marked replaced. A newer decision that clashes stands automatically. If the older one was right, record it again as a new decision.
- Look at paused work. Each handoff shows its next step. Ask someone to pick it up.
That is the whole check. No status meeting, no reading transcripts.
How do you roll it out in an hour?
Do it in this order. Each step is short.
- Agree one Campfyrr project per repository. Keep the folder name the same on every machine.
- Add a short "how we work" section to CLAUDE.md. Start from a template.
- Ask each developer to open /connect and paste the two plugin lines. A Claude Code plugin adds its hooks and MCP servers in one install (Claude Code docs).
- For cloud sessions, run the repository setup from /connect. Add the token as the
CAMPFYRR_TOKEN environment variable. Cloud sessions don't load plugins installed on your own machine (Claude Code docs). - Invite your product manager and tester. Their seats are free.
- Start one session in a shared area. Check that its brief appears.
The agents that state assumptions playbook has wording for step 2.
What doesn't Campfyrr do?
It doesn't review code, run tests or replace CI. It doesn't replace your tracker. It doesn't run or isolate sessions, so keep your worktrees. It only knows what sessions recorded. It has no audit logs yet.
What does it cost?
One person is free. Teams pay $9 per developer a month, and anyone who only reads or answers is free. Everything is free during early access. See pricing.
To bring the rest of the team in, send your product manager the page for product managers and your testers the page for QA. For why a shared session isn't enough, read sharing sessions vs. shared memory. The router problem has its own hub: stop being the human router.
The problem this answers: Sessions don't know about each other: "My teammates' Claude Code sessions don't know what each other are doing."
The principle behind this: Turn unknowns into calls: Nothing we don't know gets to stop us. We name our assumption, and we move.
Move as one at agent speed. Nobody waits. Add Campfyrr to Claude: it briefs every agent before it touches the code, keeps assumptions and handoffs where the team can see them, and never takes your code (what it keeps).