Skip to main content

For Developers

CLI workflows, SDK integration, CI/CD pipelines, and MCP tools for building agent-ready commerce.

TL;DR: Install the CLI, run check, test, and fix, then gate releases with JSON output in CI. Use MCP tools when an AI agent needs deterministic tool calls instead of scraping UI or prose.

Core Flow

npm install -g @getcolter/cli
colter check https://mystore.com --json
colter test https://mystore.com --models claude,gpt --json
colter fix https://mystore.com --dry-run --json

Apply fixes:

colter fix https://mystore.com --apply --output-dir ./colter-fixes

Key JSON Fields

FieldWhy it matters
composite_scoreRelease gate
scores.*Dimension-level failures
verdictReadiness state
fixesNext actions

CI

GitHub Actions example:

steps:
  - uses: actions/checkout@v4
  - uses: actions/setup-node@v4
    with:
      node-version: "20"
  - run: npm install -g @getcolter/cli
  - run: colter check ${{ secrets.STORE_URL }} --ci --threshold 70

Run deeper tests on a schedule:

colter test $STORE_URL --json > test-result.json

SDK

npm install @getcolter/sdk

import { ColterClient } from "@getcolter/sdk";

const client = new ColterClient({ baseUrl: "https://agenticcom.ai/api/v1" });
const result = await client.check("https://mystore.com");

MCP

Use MCP Tools when you want an agent to call:

  • colter.check
  • colter.fix
  • colter.test
  • colter.verify.run
  • commerce tools such as colter.agent.products.search

Claude Desktop example:

{
  "mcpServers": {
    "colter": {
      "command": "colter",
      "args": ["mcp"]
    }
  }
}

Evidence Packs

Generate and compare packs in CI:

colter verify https://mystore.com --out ./evidence/20260326
colter diff ./evidence/20260325 ./evidence/20260326 --json

Next Steps