Quickstart

Install specguard:

npm install -g @getcorespec/specguard

Check spec coverage on a PR:

specguard check

Setup

specguard uses the Vercel AI SDK — any supported provider works. Set the API key for your chosen provider:

# Anthropic (default)
export ANTHROPIC_API_KEY=sk-ant-...

# OpenAI
export OPENAI_API_KEY=sk-...

Override the default model (claude-haiku-4-5) via CLI flag or config file:

specguard check --model anthropic/claude-sonnet-4-20250514

Or create a .specguard.yml:

model: anthropic/claude-sonnet-4-20250514
threshold: 0.7

Pre-commit hook

Run specguard automatically before every commit to catch unspec'd changes early.

Install the hook (one-time setup per repo):

specguard hook install

This writes a .git/hooks/pre-commit script that runs specguard check --staged against your staged changes. If spec coverage is below the threshold, the commit is blocked.

To install manually, create .git/hooks/pre-commit:

#!/bin/sh
npx specguard check --staged

Then make it executable:

chmod +x .git/hooks/pre-commit

To overwrite an existing pre-commit hook:

specguard hook install --force

To remove the hook, delete .git/hooks/pre-commit.

Note: The pre-commit hook uses npx specguard, so you don't need specguard installed globally — it will be fetched from npm if needed. For faster local commits, install specguard globally: npm install -g @getcorespec/specguard.

GitHub Action

Add specguard to your CI pipeline. Set your LLM provider's API key as a repository secret:

# .github/workflows/specguard.yml
name: specguard
on: [pull_request]

jobs:
  check:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - uses: getcorespec/corespec/packages/specguard@main
        with:
          anthropic-api-key: ${{ secrets.ANTHROPIC_API_KEY }}

See it in action: demo PR with specguard failing unspec'd code

Local LLMs

You can use any OpenAI-compatible local endpoint (e.g. Ollama, LM Studio) or a self-hosted Anthropic endpoint by setting baseURL in your config file:

# .specguard.yml
model: openai/llama3.2
baseURL: http://localhost:11434/v1

For Ollama, start the server and pull a model first:

ollama serve
ollama pull llama3.2

Note: baseURL is intentionally config-file-only (not a CLI flag) to keep model configuration consolidated. All model settings (model, baseURL, and future options like temperature) live in .specguard.yml.