PulseLight × Claude Code

Pre-deploy checks inside Claude Code

PulseLight’s MCP server tells Claude Code what’s blocking launch in your repo, then hands it a paste-ready fix prompt — restraint-engineered so Claude doesn’t refactor half the codebase while it’s there.

Claude Code is Anthropic’s CLI-first agent that ships features straight into your repo.

Start free scanSee pricing

Install

Claude Code reads MCP servers from your project's .mcp.json. Once @pulselight/mcp publishes to npm, the flow below is the full install. Until then, copy fix prompts from the dashboard — same v2 prompt, no install required.

Early access

@pulselight/mcp is in early access — npm publish in flight. The snippet below previews the install once it’s live; tracking shipping at /changelog.

# Preview — once @pulselight/mcp publishes to npm:

# 1. Install the MCP server
npm install -g @pulselight/mcp

# 2. Mint a token in the dashboard:
#    app.pulselight.io → Settings → API tokens

# 3. Wire it into Claude Code — add to .mcp.json
#    at the repo root (or ~/.claude/mcp_servers.json for global):
{
  "mcpServers": {
    "pulselight": {
      "command": "pulselight-mcp",
      "env": {
        "PULSELIGHT_TOKEN": "pul_…",
        "PULSELIGHT_WORKSPACE_ID": "ws_…"
      }
    }
  }
}

How it works

  1. 1

    Ask Claude Code: is this safe to ship?

    Claude Code calls pulse_status, which returns the launch verdict, the top 3 blockers, and a per-pillar status — grounded in your real PulseLight data, not a hallucinated checklist.

    Use pulse_status to tell me what’s blocking this branch from shipping.
  2. 2

    Pick a blocker, fetch the fix prompt

    Claude Code calls pulse_top_blockers for the full founder copy + evidence, then pulse_fix_prompt to fetch the v2 restraint-engineered prompt for the chosen finding.

  3. 3

    Apply the fix, re-scan

    Claude Code applies the diff exactly as the prompt directs. Push to your branch — the PulseLight GitHub App auto-runs a fresh scan on push, and Claude Code’s next pulse_status call sees the new verdict. Green re-scan = the only proof. Repeat for the next blocker.

FAQ

Does this need a Claude Code Skill?
No — the MCP server alone is enough. Claude Code knows when to call the tools from their descriptions; you don’t need any extra orchestration. We have a procedural pulse-fix Skill in design for founders who want a single phrase trigger ("is this safe to ship?"), shipping alongside the MCP once the CLI distribution it depends on is settled.
Why MCP and not a Claude Code-specific extension?
MCP is the open protocol Anthropic, Cursor, and others have aligned on. One server works in every MCP-aware host — Claude Code today, Cursor too, future editors — without us shipping per-host extensions or per-host docs.
Does PulseLight read my code?
It scans the repo you connect via the PulseLight GitHub App. Findings stay scoped to your workspace; the share-token system is opt-in for anything public.
Do I need a paid plan?
No — Pro is enough for solo founders preparing to launch. Growth and Studio unlock more Connected Checks and projects. 14-day Studio trial included.

Ready when you are

14-day Studio trial. No card up front. Cancel any time.

Connect a repo