Getting Started

DevCortex is a structured requirements management platform built for agentic software development. It bridges the gap between human product thinking and AI-driven implementation by giving your AI agent a source of truth to work from.

• A requirements backlog with Acceptance Criteria linked to implementation
• An MCP (Model Context Protocol) server that AI agents use to read requirements and record verification results
• A CLI (dcx) for local project management
• A web dashboard to monitor agent sessions and review audit trails

Before you begin, ensure you have the following installed:

• Node.js 18 or higher
• pnpm package manager
• Code Agent: Claude Code CLI — see docs.anthropic.com/claude-code
• A DevCortex account (created in the next step)

1. Go to app.devcortexai.com/register
2. Enter your email and a secure password
3. Verify your email address via the confirmation link
4. You will be redirected to your dashboard with a default project already created

Install the DevCortex CLI globally:

npm install -g @devcortex/cli

Or with pnpm:

pnpm add -g @devcortex/cli

Verify the installation:

dcx --version

Then authenticate with your DevCortex account:

dcx login

DevCortex supports two AI coding agents. Choose the one you prefer and install it alongside the DCX CLI:

npm install -g @anthropic-ai/claude-code
dcx init
dcx start

npm install -g opencode-ai
dcx init --agent opencode
opencode

In the root directory of your software project, run:

dcx init

This creates a .devcortex-link.json file that links the local directory to your DevCortex project. Commit this file so all team members share the same project link.

{
  "orgSlug": "your-org",
  "projectId": "cmmxbv3v60008b0pz..."
}

DevCortex supports importing requirements from existing tools so you can migrate existing specs without manual re-entry.

# Preview what will be imported (no changes made)
dcx import kiro ./requirements.md --dry-run --verbose

# Import into your initialised project
dcx import kiro ./requirements.md

# Import with custom spec version and title
dcx import kiro ./requirements.md \
  --spec-version 2.0.0 \
  --spec-title "Sprint 15 Requirements" \
  --priority MUST

# Preview the import
dcx import csv ./requirements.csv --dry-run

# Import into your initialised project
dcx import csv ./requirements.csv

dcx import --list

dcx init generates the right config file automatically. Both formats connect to the same DevCortex MCP server.

{
  "mcpServers": {
    "devcortex": {
      "type": "http",
      "url": "https://api.devcortexai.com/mcp",
      "headers": { "x-api-key": "dc_live_YOUR_KEY" }
    }
  }
}

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "devcortex": {
      "type": "remote",
      "url": "https://api.devcortexai.com/mcp",
      "headers": { "x-api-key": "dc_live_YOUR_KEY" }
    }
  }
}

Your API key is available in Project Settings → MCP Connection in the DevCortex dashboard, or is generated automatically by dcx init.

With .mcp.json in place, start Code Agent from your project directory:

dcx start

Code Agent automatically loads the DevCortex MCP tools. Prompt the agent to begin implementing:

implement the next approved requirement

The agent follows the DevCortex workflow automatically:

1. Calls dc_health to verify the MCP connection
2. Fetches the next APPROVED requirement with dc_get_requirement
3. Marks it IN_PROGRESS with dc_update_requirement_status
4. Implements the code and runs tests
5. Records verification results with dc_record_verification
6. Logs the implementation with dc_log_implementation

While the agent works, use the DevCortex dashboard to track progress:

• Backlog — view requirement statuses (APPROVED → IN_PROGRESS → VERIFIED) in real time
• Activity — a full audit trail of every MCP call the agent made, timestamped and searchable
• Issues — any problems the agent logged via dc_create_issue, ready for human review
• Spec — the product specification the requirements were derived from

The recommended DevCortex development cycle:

1. Write a spec — describe your product or feature in the Spec editor
2. Generate requirements — use the AI assistant to break the spec into MUST / SHOULD / COULD requirements, each with Acceptance Criteria
3. Approve requirements — review and set status to APPROVED for the requirements you want the agent to work on
4. Run an agent session — start Code Agent and let it implement the approved requirements
5. Review and verify — check the implementation against the ACs, mark VERIFIED when satisfied
6. Repeat — add more requirements to the backlog and start the next sprint