Skip to main content
This page focuses on developing and running ECP locally. It does not cover how this documentation site is built.

​
Prerequisites

  • Node.js 22+ (required for schema generation and CI)
  • npm (v7+ for workspaces) or pnpm
  • For OpenAI: set OPENAI_API_KEY
  • For Ollama: install Ollama and run it locally

​
Install from npm (published packages)

npm install -g @executioncontrolprotocol/cli
ecp validate spec.yaml --input shopifyStoreId=test --input jiraProject=TEST

​
Build and use the CLI from the repo

From the repository root: 
npm install
npm run build
npm run build runs a full clean of package dist/ outputs first, then compiles (see the repo root package.json). Then either use the workspace binary: 
npx ecp validate spec.yaml --input shopifyStoreId=test --input jiraProject=TEST
Or link the CLI globally: 
cd packages/cli && npm link && cd ../..
(ecp runs compiled dist/index.js via bin/ecp.cjs.)

​
Environment and config

  • OpenAI: Set OPENAI_API_KEY.
  • Ollama: Default base URL is http://localhost:11434; override with --ollama-base-url or OLLAMA_BASE_URL.
  • System config: Optional ecp.config.yaml (e.g. from config/ecp.config.example.yaml) with version: "0.5", wiring under models, tools, loggers, …, and policy under the security mirror. The CLI loads it from --config, ./ecp.config.yaml, or ~/.ecp/config.yaml.
See SETUP.md for full details.

​
Run examples

From the repo root (after npm run build): 
ecp run examples/single-executor/context.yaml --enable openai -i topic="ECP"
ecp validate examples/single-executor/context.yaml -i topic=test
To inspect what happened during a run, use ecp trace (tracing is enabled by default; disable with --no-trace). For the full, up-to-date flag list, run ecp --help or ecp <command> --help (for example ecp run --help).

​
Tests

Unit tests expect built packages (dist/). From the root: 
npm run build
npm run test:unit
Or npm test (builds first, then unit + integration + e2e).