Campfire Quick Start

Install Campfire, record yourself, and start the local API and Viewer.

Install Campfire, run it once to record the human, then start the local API and Viewer. The agent creates the workspace and goal when work starts.

Install

Campfire requires Node.js 22 or newer.

Choose one install method:

# curl installer
curl -fsSL https://boringinfra.company/campfire/install.sh | sh

# Homebrew
brew install boringinfraco/campfire/campfire

# Nix
nix profile install github:BoringInfraCo/Campfire

The curl installer puts the binary in $CAMPREFIX/bin. The default prefix is /usr/local when that directory is writable, and ~/.local otherwise.

First run

You shouldn’t have to design a multi-agent system just to get two agents working on the same problem.

campfire
campfire up

campfire records you once. The suggested name comes from git config user.name or your login. It does not ask for an agent, a workspace, or a goal. Credentials are stored under ~/.config/campfire.

campfire up starts the local API and the token-free Viewer in this process, and connects each installed Codex or OpenCode harness. It does not import past sessions. Registering an agent session stays an explicit step.

Then look at what’s happening:

campfire status

You’ll see the current workspace without credentials. Nobody has to reconstruct that from three different agent sessions.

Scripts can create the first human, agent, workspace, and goal in one shot. That path does not start the server. --json prints each credential once:

campfire onboard \
  --human-name "Sergio" \
  --agent-name "Codex" \
  --harness codex \
  --workspace-name "billing deploy" \
  --goal "Ship the billing migration safely" \
  --json

Keep the work when a session ends

An agent session ends. The work shouldn’t disappear with it.

campfire status

The next agent reads the same workspace: the goal, what happened, what was decided, and what’s left.

Next steps

Overview

Use Cases