$ man how-to/claude-code-co-work

A co-work session is a Claude Code instance pointed at a shared folder. Everyone on the team gets the same context: the same CLAUDE.md rules, the same skills, the same data files, the same voice system. It is not a chat room. It is a shared operating environment where the AI has already read your playbooks. The difference from traditional collaboration: Notion docs sit in a browser tab nobody has open. A Confluence page gets written once and forgotten. A co-work folder is active. The AI reads it on every session start. Every workflow is executable, not just documented. Every playbook runs, it does not just explain.

PATTERN

Drop a folder with: CLAUDE.md (team rules, voice, anti-slop filters), scripts/ (enrichment crons, lead scoring, campaign automation), skills/ (outreach generation, research compilation, pipeline review), data/ (target accounts, enrichment results, campaign metrics). New SDR joins. They open Claude Code in the folder. The CLAUDE.md loads automatically. They say "research Acme Corp for outbound" and the agent runs the research skill, pulls from Exa, enriches through Apollo, checks Attio for existing history, and produces a research brief. The SDR did not read a single doc. The folder was the onboarding. This is the active playbook model. Documentation that executes. Context that compounds. Every session builds on previous sessions through handoff files and memory.

CODE

Create a repo. Structure it: CLAUDE.md - team rules, model preferences, voice system .claude/skills/ - one skill per workflow (research, outreach, reporting) scripts/ - automation scripts that skills invoke data/ - shared data files (target lists, enrichment results) tasks/ - task tracking and lessons learned Each team member clones the repo and runs Claude Code from the root. The CLAUDE.md establishes shared context. Skills provide consistent workflows. Data files keep everyone working from the same source of truth. The key insight: the repo IS the team knowledge base. Not a separate wiki. Not a Notion workspace. The same folder that runs the automation also documents the automation. Code and documentation live together.

PRO TIP

Passive doc: "To research a prospect, check LinkedIn, look at their recent funding, review their tech stack on BuiltWith, and summarize findings in a Google Doc." Active playbook: a skill file that Claude Code reads and executes. It checks LinkedIn via the browser, pulls funding data from Exa, queries BuiltWith through the API, and writes the summary to the research folder in the repo. Same workflow. One requires a human to follow steps. The other requires a human to say "research this company." Every passive doc in your org is a candidate for an active playbook. The conversion process: identify the steps, write them as a skill, test the skill, deploy to the shared folder. The documentation becomes the automation.