Getting Started

From zero to acceptance in just three steps.

---

Installation

npm install -g speclore

Prerequisites

SpecLore requires Node.js 18+ and any of pnpm / npm / yarn package managers.

Initialize Your Project

cd your-project && speclore setup

speclore setup will:

1. Detect AI tools: Automatically scan for Cursor / Claude Code / Qoder in your project
2. Write MCP config: Generate MCP configuration files for detected AI tools
3. Generate rules: Create .speclore/config.yaml core configuration

✔ Detected AI tools: Cursor, Qoder
✔ Written .cursor/mcp.json
✔ Written .qoder-cn/mcp.json
✔ Generated .speclore/config.yaml

Generate Acceptance Criteria

Generate standard BDD .feature files from any requirement source format:

# From a Markdown file
speclore spec requirements.md

# From plain text
speclore spec "Patient registration requires phone verification and WeChat login"

# From a Word document
speclore spec design.docx

# From a URL
speclore spec https://jira.example.com/issue/PROJ-123

Example output:

✔ Generated specs/patient/register.feature
  Scenario 1: Register with valid phone number
  Scenario 2: Reject invalid phone format
  Scenario 3: Warn on duplicate phone number

Generate AI Coding Constraints

speclore code

Automatically generates constraint files in the correct format for detected AI tools:

AI Tool	Constraint File
Cursor	.cursor/rules/speclore.mdc
Claude Code	.claude/rules/speclore.md
Qoder	.qoder/rules/speclore.md

Also generates test scaffolding files (e.g., tests/patient/register.test.ts) with it.skip test cases mapped to .feature scenarios.

Code & Acceptance

Code normally in your AI client. The AI will automatically read the constraint rules while you fill in the it.skip test scaffolding.

When ready, run acceptance:

speclore verify

Example output:

✔ 3/3 scenarios passed (100%)

specs/patient/register.feature
  ✓ Register with valid phone number    → passed
  ✓ Reject invalid phone format         → passed
  ✓ Warn on duplicate phone number      → passed

✅ Acceptance passed

Using in AI Clients

speclore setup has already configured MCP automatically. You can use natural language in your AI client to complete the entire workflow:

You: Help me implement patient registration with phone verification

AI: Generated specs/patient/register.feature with 3 acceptance scenarios...

AI: Generated coding constraints and test scaffolding...

You: OK, I'll implement the code and tests

You: Run acceptance

AI: ✅ 3/3 scenarios passed (100%)

Next Steps

• Learn the full Workflow
• Check the Configuration Reference
• Explore MCP Tools in detail