Skip to content

Installation

Foundry installs into your project: skills go into your AI tool's skill directory, support files go under .foundry/. The installer is an interactive wizard — it asks who you are, where to write output, which modules to include, and which AI tool(s) you use. Nothing is installed system-wide.

Prerequisites

  • Node.js >= 20. Check with node --version.
  • A GitHub personal access token with read:packages scope. Foundry is published to GitHub Packages, not the public npm registry. Create the token at https://github.com/settings/tokens if you don't have one.

Configure npm for GitHub Packages

Add this block to your project's .npmrc (or ~/.npmrc for global access across projects):

ini
@engineering11:registry=https://npm.pkg.github.com
//npm.pkg.github.com/:_authToken=YOUR_GITHUB_TOKEN

401 Unauthorized when installing? Almost always a missing or mis-scoped token. See Troubleshooting.

Install

From your project root:

bash
npx @engineering11/e11-foundry install

The wizard collects, in order:

PromptDefaultWhat it controls
What should agents call you?E11 UserAgent salutations and document headers
Communication languageEnglishLanguage agents use when chatting with you
Document output languageEnglishLanguage used in produced artifacts (PRDs, stories, …)
Output folder.foundry-outputRoot for all generated artifacts
Project namecurrent directory nameUsed in document headers and config
Development experienceintermediatebeginner / intermediate / expert — controls how verbose agent explanations are
Planning artifacts folder{output_folder}/planning-artifactsBriefs, PRDs, UX, architecture, epics
Implementation artifacts folder{output_folder}/implementation-artifactsSprint status, stories, reviews, retros, quick-flow specs
Long-term project knowledgedocsResearch, references, document-project output
Modulesfoundry (preselected)Which skill catalogs get installed
Tools / IDEsclaude-code, cursor (preferred)Which of the supported tools get skill directories

All of these can be passed as CLI flags too (--user-name, --communication-language, --document-output-language, --output-folder, --modules, --tools, --directory, -y); see npx @engineering11/e11-foundry install --help.

Source of these settings: src/foundry/module.yaml. Your answers land in central config — team-scope answers in .foundry/config.toml, personal ones (name, languages, skill level) in .foundry/config.user.toml.

Install the knowledge corpus

Foundry's E11-aware agents look up SDK docs and patterns through @engineering11/alloy. Install it in the same project:

bash
npm install @engineering11/alloy

The skill:e11-knowledge resolves each query through a ladder (MCP tools, the pinned cache, then node_modules/@engineering11/alloy/). Without a corpus reachable on any rung, any agent that consults alloy halts with a per-rung fix. You don't have to remember the prerequisites: the installer actively wires alloy - it installs the corpus, pins .alloy/corpus.lock, provisions the CLI, syncs the cache, and registers the alloy-corpus MCP server - and its completion banner prints the exact fix for any genuine offline/registry-down/unpublished case.

What gets written

After install completes:

text
.claude/skills/                # (and/or .agents/skills/, .opencode/skills/ per your tool picks)
  e11-master/                  #   one directory per installed skill — agents,
  e11-agent-dev/               #   lifecycle workflows, and anytime tools alike
  e11-prd/
  ...
.foundry/                      # Support files, not skills
  config.toml                  #   central config, team scope (installer-owned)
  config.user.toml             #   central config, user scope (gitignored)
  custom/                      #   YOUR overrides — never touched by the installer
  scripts/                     #   resolve-customization.mjs, resolve-config.mjs, …
  _config/e11-help.csv         #   assembled catalog of every installed skill
  foundry/                     #   the installed module content + runtime config.yaml
.foundry-output/               # (or whatever output_folder you chose)
  planning-artifacts/
  implementation-artifacts/
docs/                          # (or whatever project_knowledge you chose)

Every agent, workflow, and task installs the same way — as a skill directory your tool reads natively. See Tool integrations for the per-tool details.

Update an existing installation

bash
npx @engineering11/e11-foundry install

The installer detects the existing install and updates it in place: modules are reinstalled at their newest version, stale or renamed skill directories are cleaned from your tool dirs, and your config answers and everything under .foundry/custom/ are preserved. Pass --action quick-update to skip re-prompting and only answer questions for newly added settings.

If your install predates agents-as-skills, the same command runs the migration — see Upgrading Foundry.

Check status

bash
npx @engineering11/e11-foundry status

Prints installed modules and versions, plus available updates.

Uninstall

bash
npx @engineering11/e11-foundry uninstall

Removes .foundry/ and tool-specific files. Preserves your generated artifacts (.foundry-output/, docs/).

Next

First run — what got installed, how to invoke your first skill, and how much ceremony to dial in.

Proprietary to Engineering11 LLC.