Drydock – Agile/TDD Spec Driven Delivery – Initial Release Web Cloud Studio released Drydock, an open-source Python CLI tool that adds an Agile/TDD process layer to specification-driven development using LLMs like Claude and Codex. The tool enables developers to import source material, analyze it into stories, review plans, and build software iteratively without per-token API billing. Drydock is available under the MIT License and targets subscription-authenticated CLI users. The missing process layer for specification-driven development. Drydock gives specification-driven development the part it skipped: a repeatable Agile delivery and Test Driven Development based end-to-end process that turns specs into reviewed plans, context-managed builds, and obvious ways to maintain. Install install · 60-Second Example 60-second-example · Why It Is Different why-it-is-different · Canonical Specification https://webcloudstudio.com/project-docs/drydock/ Overview Deck https://webcloudstudio.com/drydock/ · White Paper https://zenodo.org/records/21287574 · 10 Minute Overview https://webcloudstudio.com/project-docs/drydock/presentation/Drydock Video.web.mp4 · User Installation Guide https://webcloudstudio.github.io/Drydock/USER INSTALLATION.html uv tool install drydock-sdd then: drydock init MyApp && drydock status Drydock runs on your existing Claude or Codex subscription CLI. It does not require API keys or per-token API billing. Drydock adds the missing process layer to specification-driven development: import source material, analyze it into stories and acceptance criteria, review decisions in the QuarterDeck, plan a dependency graph, build one context-optimized step at a time, verify evidence, and keep the Blueprint and software aligned as the product changes. You are the Commander— the product owner. The LLM is your Agile delivery team. Drydock makes that relationship explicit, reviewable, and repeatable. Copyright c 2026 Web Cloud Studio. Licensed under the MIT License. See LICENSE /webcloudstudio/Drydock/blob/main/LICENSE . uv tool install drydock-sdd drydock config set llm provider claude drydock init MyApp --description "A small web application." drydock import MyApp ./notes --format markdown drydock analyze MyApp drydock run quarterdeck MyApp drydock plan MyApp drydock run quarterdeck MyApp drydock build MyApp Edit Some Blueprints drydock refit MyApp drydock build MyApp That loop creates a Target workspace, decomposes the source material, opens the Commander review surface, builds a dependency graph, and executes the first runnable frontier with persisted evidence. Drydock is an installable Python command-line package. The PyPI distribution is drydock-sdd ; the installed command is drydock . Drydock implements the SAIL methodology: | Phase | Purpose | Primary commands | |---|---|---| | Set Up | Install, configure, and initialize a Target workspace | config , init , status | | Analyze | Import material and decompose it into stories, blockers, and acceptance milestones | import , analyze , run quarterdeck , plan | | Implement | Build the Manifest frontier and verify evidence | build , build status , build verify , rigging , document | | Loop | Manage change while preserving the Blueprint as source of truth | refit , build , document | The core idea is simple: reproducible LLM builds require a process. Drydock uses Agile structure, explicit product-owner review, durable evidence, and context-managed build prompts so generated software can be inspected, repeated, and iterated. Drydock is for subscription-authenticated CLI users. It does not use API-key-backed model calls and does not require per-token API billing. LLM-assisted commands execute through a locally authenticated provider CLI: claude for Anthropic Claude subscription CLI users. codex for OpenAI Codex subscription CLI users. Set the provider with: drydock config set llm provider claude or drydock config set llm provider codex The provider CLI must already be installed, authenticated, and available on PATH . Deterministic commands such as status , validate , document assemble , and publish do not call an LLM. Python 3.11 or later is required. Recommended: uv tool install drydock-sdd Alternative: pipx install drydock-sdd Virtual environment install: python -m pip install drydock-sdd Verify: drydock --version drydock --help PDF publishing drydock publish --pdf is optional and requires the pdf extra plus a local Chromium download: uv tool install "drydock-sdd pdf " playwright install chromium See docs/USER INSTALLATION.md /webcloudstudio/Drydock/blob/main/docs/USER INSTALLATION.md for the full installation guide. Drydock ships two Claude Code skills that support the Loop phase: /refit design discussion captured to a notes file in the Target directory and /apply-refit turn the captured decisions into change tickets . Installing them is optional: python python -c "import shutil, pathlib; from drydock.paths import get rigging root; \ dest = pathlib.Path.home / '.claude' / 'skills'; \ shutil.copytree s, dest / s.name, dirs exist ok=True \ for s in get rigging root / 'skills' .iterdir " The skills are copied from the packaged Rigging/skills/ directory into ~/.claude/skills/ . See docs/Drydock SKILLS.md /webcloudstudio/Drydock/blob/main/docs/Drydock SKILLS.md for usage. Create one projects directory, configure Drydock's workspace and build output root, and initialize a Target. export PROJECTS="$HOME/projects" mkdir -p "$PROJECTS/drydock" drydock config set drydock workspace "$PROJECTS/drydock" drydock config set drydock build directory "$PROJECTS" drydock config set llm provider claude drydock init MyApp --display-name "My App" --description "A working software product." drydock status Import source material and run the planning loop: python drydock import MyApp ./notes --format markdown drydock analyze MyApp drydock run quarterdeck MyApp drydock plan MyApp drydock build status MyApp Build and verify one frontier at a time: drydock build MyApp drydock build status MyApp drydock build verify MyApp