Agentic-first environment variable manager. No UI, no SDK — the AI agent is the user. The API is the product.
# Build
make build
# Run (zero config)
./envkit
# Or with custom port/data
./envkit --port 9090 --data ./envkit.json# 1. Request OTP
curl -X POST http://localhost:8080/auth/request -d '[email protected]'
# 2. Verify OTP (check server logs for the code)
curl -X POST http://localhost:8080/auth/verify -d '[email protected]&code=123456'
# Returns: token=env_xxxxx workspace=ws_xxxxx
# 3. Use the token
curl -H "Authorization: Bearer env_xxxxx" http://localhost:8080/projects# Create
curl -H "Authorization: Bearer $TOKEN" -X POST http://localhost:8080/projects -d 'name=myapp'
# List
curl -H "Authorization: Bearer $TOKEN" http://localhost:8080/projects
# Get
curl -H "Authorization: Bearer $TOKEN" http://localhost:8080/projects/proj_abc123
# Delete
curl -H "Authorization: Bearer $TOKEN" -X DELETE http://localhost:8080/projects/proj_abc123# Create
curl -H "Authorization: Bearer $TOKEN" -X POST http://localhost:8080/environments -d 'project=proj_abc123&name=production'
# List by project
curl -H "Authorization: Bearer $TOKEN" 'http://localhost:8080/environments?project=proj_abc123'
# Delete
curl -H "Authorization: Bearer $TOKEN" -X DELETE http://localhost:8080/environments/env_xyz789# Create
curl -H "Authorization: Bearer $TOKEN" -X POST http://localhost:8080/variables -d 'environment=env_xyz789&key=DATABASE_URL&value=postgres://localhost&secret=true'
# List by environment
curl -H "Authorization: Bearer $TOKEN" 'http://localhost:8080/variables?environment=env_xyz789'
# Delete
curl -H "Authorization: Bearer $TOKEN" -X DELETE http://localhost:8080/variables/var_mno456# .env format
curl -H "Authorization: Bearer $TOKEN" http://localhost:8080/export/env_xyz789?format=env
# Shell format
curl -H "Authorization: Bearer $TOKEN" http://localhost:8080/export/env_xyz789?format=shellcurl -H "Authorization: Bearer $TOKEN" http://localhost:8080/auditAdd Accept: application/json header or ?format=json query param for JSON responses.
The service speaks Model Context Protocol at /mcp for chat client integrations.
GET /help— one-page operating manualGET /.well-known/agent.md— agent-friendly documentation
- Language: Go
- Storage: JSON file (zero external dependencies)
- Architecture: Single binary, internal package structure
- Build:
make build,make test,make vet
MIT