Your AI assistant answers from memory. Now it can answer from a measurement of the culture, current to the hour.

Moodlight is a real-time cultural reasoning engine. Every hour it measures live news and social conversation, scores it for emotion, and tracks which conversations are accelerating, which are crowded, and which are still open. The Moodlight MCP server lets the assistant you already use, Claude, Claude Code, Cursor, or any MCP-compatible tool, draw on that measurement, and on Moodlight's judgment, in the middle of your own work.

Nothing to download. Subscribe, a key arrives by email, add it as a connector, and your assistant gains seven new tools.

What your assistant gains

Measurement tier: the substrate itself. Your model does the reasoning.

category_signal: the raw measurement for a tracked category. Coverage, composition, emotional register, accelerating or cooling against the whole corpus. About a second.
brand_signal: a brand's genuine tracked mentions over the last seven days, namesakes filtered, emotional register, topic mix. About a second.

Judgment tier: the measurement plus Moodlight's own read.

start_read / get_read: a full strategic read on a brand, category, or question. The call, what would make it wrong, who it is for. Each call returns in a second; the read takes 60 to 90 seconds.
cultural_read: the same read in one call, for clients that can wait.

Agents tier: the Moodlight agent marketplace, inside your assistant.

list_agents: 32 strategy, creative, growth, and validation agents, Brand Auditor to David Ogilvy, each grounded in the live measurement.
run_agent: run one on a brief. Returns in a second; the agent takes two to seven minutes. Agents chain: pass an earlier job as upstream and the next one builds on it. "Run the Brand Auditor on Inuvo, then David Ogilvy on top of it" is a complete prompt.

Pricing

Per organization. Every plan includes a 30-day trial and the Measurement tier. Annual is ten months. Founding prices are locked for twelve months for the first twenty organizations.

Measurement

$250/month

The raw signal. category_signal and brand_signal.

Monthly $2,500/yr

Judgment

$750/month

Measurement plus full strategic reads, 150 a month.

Monthly $7,500/yr

Agents

$1,500/month

Everything, including all 32 agents, 75 runs a month.

Monthly $15,000/yr

Card required for the trial; nothing is charged for 30 days and you can cancel any time before that. Your key and connection steps arrive by email within a minute of checkout.

Connecting

Claude (desktop, web, phone): Settings, Connectors, Add custom connector. Name it Moodlight, URL https://mcp.moodlightintel.com/mcp, Add. Paste your key on the Moodlight page and press Allow. Connect once on the web and it is connected on your phone too.

Claude Code:

claude mcp add --transport http --scope user moodlight https://mcp.moodlightintel.com/mcp --header "Authorization: Bearer YOUR_KEY"

Cursor: add to ~/.cursor/mcp.json

{"mcpServers": {"moodlight": {"url": "https://mcp.moodlightintel.com/mcp",
  "headers": {"Authorization": "Bearer YOUR_KEY"}}}}

Then ask it something hard.

What Moodlight sees, and what it never sees

Moodlight never has access to your files, documents, conversation history, or any other tool your assistant uses. It receives only the question or brief your assistant sends to a Moodlight tool.
Briefs and outputs are kept for 30 days so a finished job can be collected, then deleted. Keys are stored as one-way hashes. No profile, no memory between requests, nothing used to train anything.
A read is Moodlight's judgment, not advice. Every read names the condition under which it would be wrong. Check it before you act on it.
The measurement is a tracked sample, not a census. A brand that is quiet in the signal has low news salience, not low cultural presence, and Moodlight says so when it applies.
Early access: live and monitored, no formal uptime guarantee yet.