Claude

This workflow uses Claude Projects to onboard a new engineer to an existing repository — a common pain point where context volume defeats normal chat limits.

Repo onboarding with Projects + Artifacts

1. Create a Project. Name it after the repo. Upload architecture docs, README, ADRs, and the top-level directory tree export (sanitized of secrets).
2. Ask for a map. Prompt: "Summarize module boundaries, entry points, test layout, and the three riskiest areas for new contributors." Claude uses Project knowledge across sessions.
3. Generate an Artifact. Request a living onboarding doc in Markdown — setup steps, env vars (names only), and a two-week learning path. Edit inline in the Artifact panel.
4. Drill into files. Paste specific file paths or chunks for deep dives. Ask for sequence diagrams or dependency notes; keep each thread focused on one subsystem.
5. Hand off. Export the Artifact to your wiki. New hires start in the Project instead of re-uploading the same PDFs.

Security note: never upload credentials, customer data, or unredacted production logs — treat Projects like a shared drive with model access.

Claude's free tier offers limited daily messages on current Sonnet-class models — enough to evaluate tone and context handling, not enough for all-day professional use. Limits reset on a rolling basis and can tighten when Anthropic adjusts capacity.

Claude Pro (typically around $20/month) raises usage caps, unlocks higher-end models (including Opus-tier where available), priority at high traffic, and full Projects/Artifacts workflows. Team and Enterprise plans add seat management, billing consolidation, and organizational policies — check Anthropic for current seat minimums and data handling terms.

The Anthropic API bills per token separately from Pro. API model IDs, context pricing, and batch discounts change frequently; Pro subscription does not include API credits. Confirm model names, rate limits, and regional availability on Anthropic's official site before production deployment.

ChatGPT remains the better default when you want Custom GPTs, wider third-party integrations, voice-first mobile use, or OpenAI's latest reasoning models for mixed media tasks. Choose ChatGPT if your org already standardized on OpenAI API keys.

Gemini fits Google-centric teams: Workspace sidebars, Android system integration, and Deep Research flows that pull from Search and YouTube. If your sources live in Drive and Gmail, Gemini reduces copy-paste friction Claude cannot match.

Cursor overlaps on coding but solves a different problem — IDE-embedded multi-file edits and agentic refactors. Many developers pair Claude Pro for architecture discussions with Cursor for implementation inside the repo.