seaLeye
Bring your app. sealeye run reads it, asks a few questions in your language, writes a
deployment plan you can read, and on "yes" builds a real environment and hands you a
report. When your app needs something the platform cannot build, it says so, with the
code that shows it, and builds nothing.
- seaLeye.run is the
sealeye command.
- seaLeye.online is the web portal (
portal/): a new environment from a folder,
each environment's state, history, logs, evidence, plan and report and tear down, a
plan's check, the catalog (with which apps use each module) and the needs ledger,
the team and costs, and health beside your name; for an organization-wide token, a
Platform section with discovery (every catalog with the run that wrote it), the
platform values and the gap report.
- Both are clients of the seaLeye API in
api/, which runs at https://api.sealeye.online.
Use it
curl -fsSL https://sealeye.run/install | sh
sealeye login # paste the token your platform team gave you
cd your-app && sealeye run
The command talks to https://api.sealeye.online unless SEALEYE_API (in
~/.sealeye/config) or --api says otherwise.
All the documentation starts at docs/README.md: where it is going
(VISION), what is next (ROADMAP), and how it is
built (ARCHITECTURE).
Run it locally
You need Node 20 or later, zip, the AWS CLI signed in to an account, and an Anthropic
API key. gitleaks on your PATH adds the API's deeper secret scan, and OpenTofu 1.12.6
is needed only to build an organization's own Terraform modules.
npm install
npm test # no AWS, no API key needed
npm link # puts `sealeye` on your path
Settings go in .env at the repo root (ignored by git):
export SEALEYE_ACCOUNT=<the AWS account id you expect to be signed in to>
export SEALEYE_BUCKET=<an S3 bucket for build artifacts, in ca-central-1>
export SEALEYE_LWA_VERSION=<the Lambda Web Adapter layer version>
export SEALEYE_FUNCTION_BOUNDARY_ARN=<the deploy-role stack's FunctionBoundaryArn output (deploy/README.md)>
The AWS credentials must also be able to read /sealeye-apps/ in Parameter Store, where
an app's settings are kept.
Your own API key goes in your environment or in ~/.sealeye/config:
ANTHROPIC_API_KEY=...
ORG=Your Organization
Start the API in one window, then use the command in another, pointed at it (add
SEALEYE_API=http://localhost:3000 to the repo .env to make that the default here).
The local API asks for a token too: node scripts/tokens.js add [email protected] "Your Name" prints one for sealeye login --api http://localhost:3000, or start it with
SEALEYE_AUTH=off npm start to skip that on your own machine.
npm start
cd apps/beneficiary-lookup-api && sealeye run --api http://localhost:3000
sealeye doctor --api http://localhost:3000 # checks every one of the above
Settings and files from before 2.17.0 (DEPLOYME_*, ~/.deployme/, deployme.yaml)
are still read.
Layout
| Folder | What is in it |
|---|
cli/ | the sealeye command and its interview |
api/ | the API and the executor that builds environments |
core/ | the plan contract and settings both of them share |
prompts/, templates/ | what the model is told, and the CloudFormation the starter's modules are built from (an organization's own modules are built with OpenTofu, api/executor/terraform/) |
portal/ | seaLeye.online, with its own package.json: cd portal && npm ci && npm run build, then npm run test:e2e |
apps/, plans/ | sample apps and test plans |
test/ | npm test |
scripts/ | the evals (npm run eval:interview, npm run eval:discovery) and the server's one-off jobs |
deploy/ | the pilot server and its runbook, deploy/README.md |
sites/ | seaLeye.run (with the installer) and sealeye.org |
docs/ | everything else, from docs/README.md |
brand/ | the brand for both surfaces |
archive/ | the October 2026 demo, read-only |
Rules for working in this repo are in CLAUDE.md; decisions and their
reasons in DESIGN.md.
