# Bringing AIWG into an Existing Project You already have a codebase. Maybe it's yours, maybe you inherited it. Either way, you want an AI assistant that actually understands what's in there — not one that has to be reminded every session what the project does and how it's structured. This guide gets AIWG loaded and oriented in under ten minutes. --- ## Step 1 — Install ```bash npm install -g aiwg ``` --- ## Step 2 — Establish project setup ```bash cd /path/to/your/project ``` Ask your AI assistant to set up AIWG for this repository: ``` Help me set up this project for AIWG. ``` The setup conversation should establish how this repository actually works: which remotes matter, where issues live, whether delivery is direct-to-main or branch/PR based, what commit and signing policy applies, which providers should be enabled, and whether `.aiwg/issues/` should be used for local issue storage. The assistant may call `aiwg setup project --dry-run` or `aiwg setup project --yes` while working with you, but the important surface is the guided conversation. You should not have to map repo policy fields by hand. --- ## Step 3 — Deploy to your project ```bash aiwg use sdlc ``` This installs the agents, commands, and rules into `.claude/`. Claude Code picks them up automatically on the next session. --- ## Step 4 — Let the AI read the codebase Open Claude Code: ```bash claude . ``` Then run the codebase intake scan: ``` /intake-from-codebase ``` This scans your project — directory structure, key files, existing tests, README, package.json, config files — and generates a project intake document in `.aiwg/intake/`. It figures out: - What the project does - What tech stack it uses - What the existing architecture looks like - What tests exist (and what's missing) You don't have to answer questions or fill out forms. It reads the code. --- ## Step 5 — Review what it found The intake document lands at `.aiwg/intake/`. Read it and correct anything that's wrong: ``` What did you learn about this project? ``` Or open `.aiwg/intake/project-intake.md` directly and edit it. This document drives everything else — fix inaccuracies now. --- ## Step 6 — Set up project context After the repo policy setup and intake scan, ask the assistant to wire the project context so future sessions load the right background: ``` /aiwg-setup-project ``` This wires the intake into the provider configuration so the AI has persistent context across sessions. It complements the earlier setup conversation about remotes, issue storage, delivery behavior, and policy. --- ## Now the AI knows your project Ask it anything: ``` What are the riskiest parts of this codebase? ``` ``` Find functions with no test coverage ``` ``` Where does the authentication flow start and what does it do? ``` ``` What would break first if we changed the user model? ``` ``` Summarize the architecture for a new engineer joining the team ``` The AI answers as a team member who has read all the code — not as a generic assistant starting from zero. --- ## If the project has no documentation That's a common situation. AIWG can generate it: ``` Write architecture documentation based on what you've read ``` ``` Create a README from what you know about this project ``` ``` Generate use cases from the existing code ``` The resulting docs go into `.aiwg/` and can be committed as part of the project. --- ## If you inherited code you don't fully understand This is exactly what AIWG is built for. Run: ``` Explain this codebase to me as if I just joined the team today ``` Then dig into specific areas: ``` What does the billing module do and how does it connect to the rest of the system? ``` ``` Walk me through what happens when a user logs in ``` ``` Which parts of this code look like technical debt? ``` --- ## Ongoing use Once the project is loaded, AIWG is most useful in your normal workflow: - Ask the agent for project status at the start of each session ("where are we?", "what's next?") - For iterative tasks, ask the agent to keep working until a measurable check passes ("fix all failing tests; keep iterating until `npm test` passes") - Ask for a security review periodically as a quality gate ("run a security review on this branch") The agent invokes the underlying AIWG workflows — `/project-status`, agent loops, `/flow-security-review-cycle` — from inside the conversation. You stay in chat. See [New Project](new-project.md) for a full description of the SDLC workflow once you're oriented.