There’s a pattern that keeps repeating in software. A powerful data layer gets locked inside a single interface. Then someone figures out how to break it open, and everything changes. It happened with content. It happened with commerce. And now it’s happening with buyer intelligence.
For years, teams have accessed their buyer data—signals, enrichment, contact records, and account context—through a single interface. Call it the CRM, the platform, the dashboard. That data was powerful, sure. But it wasn’t portable. It lived inside the tool.
AI agents changed the equation.
Suddenly, you need your buyer intelligence to be available not just in a UI, but in a pipeline. In a terminal. In a chat window. Inside Claude, ChatGPT, Copilot, Cursor. Inside the custom agent your RevOps team is already building.
Buyer intelligence needs to be headless.
Today, that's exactly what we launched.
Introducing headless buyer intelligence
Common Room is launching two new execution surfaces that take buyer intelligence out of the interface and make it programmable infrastructure:
1. Common Room CLI (cr) — Headless by design
The cr CLI is a fully scriptable, headless interface to the Common Room API. No browser required. No UI to navigate. Just direct, structured access to your buyer intelligence from the terminal, a CI/CD pipeline, or inside any custom AI agent.
This is buyer intelligence as infrastructure. The same identity-resolved, continuously enriched, signal-unified data that powers the Common Room platform, now accessible headlessly from anywhere you build.
Key capabilities:
- cr object list contact — query any object with typed filters and get exactly the records you need
- cr contact create / cr activity create — write data back with full CRUD; no UI, no manual entry
- Multiple output formats (table and JSON) — pipe to any downstream tool or LLM
- Agent-context command — emits a machine-readable map of the entire CLI surface to inject into any agent’s system prompt
- Three auth modes — environment variable tokens for CI/CD, browser OAuth for interactive use
2. MCP Server Write Capabilities: Conversational AI, now with action
The Common Room MCP Server already lets AI assistants like Claude, ChatGPT, and Copilot read your buyer intelligence. Now they can act on it.
- New write capabilities mean AI assistants can actually do things, not just surface information. Create and update contacts, log meetings and activities, manage segments. All without leaving the conversation.
The same governance and permissions that apply to any Common Room user apply here. Admins control whether write tools are enabled.
Why headless matters for AI-native GTM
AI agents don't use dashboards. They don't click through menus. They call tools, retrieve context, and execute actions programmatically at scale.
The problem: most buyer intelligence still lives inside UI-first platforms. If your data isn't accessible headlessly, your agents are operating blind, or stitching together brittle integrations that break.
That's the gap the Common Room CLI and MCP Server closes.
Both surfaces expose the same underlying intelligence layer—identity-resolved buyer data, continuously enriched by DataAgent, unified across first-, second-, and third-party signals—as headless infrastructure that AI systems can operate on directly.
The companies that win with AI won't simply have access to better models. They'll have headless buyer intelligence that any AI agent can run on top of.
Two surfaces. One intelligence layer.
The CLI and MCP Server are complementary. Both are headless. The difference is the interaction model.
CLI: Deterministic headless execution
The CLI is for when you need control. RevOps and GTM engineers use it for automation pipelines, scheduled jobs, batch operations, and custom AI agent backends. You define the filters, the output format, the retrieval logic, and the downstream workflow. Predictable. Repeatable. Production-grade.
MCP: Conversational headless execution
The MCP Server is for when you need flexibility. AI assistants like Claude, ChatGPT, Copilot, and Cursor dynamically determine what context is relevant, which tools to call, and what actions to take during a conversation. A rep says, "What's happening with Acme?" and the assistant pulls live buyer intelligence automatically. They say, "Log my meeting with Sarah," and the write tool fires.
Most organizations will use both. CLI for production automation. MCP for humans using AI assistants every day.
What headless buyer intelligence unlocks
FOR REVOPS & GTM ENGINEERS
- Build agentic pipelines that sync contacts, refresh segments, and log activities—fully automated
- Run nightly data hygiene, weekly enrichment checks, batch CSV exports of filtered contact lists
- Use cr as the tool layer in any custom AI agent—structured, typed responses, zero boilerplate
- Inject agent-context into your agent's system prompt for instant CLI awareness
FOR AI ASSISTANTS (CLAUDE, CHATGPT, COPILOT, CURSOR)
- "What's happening with Acme Corp?" Assistant pulls live activities, contacts, and engagement signals
- "Log my meeting with Sarah at Acme" MCP write tool creates the activity record directly
- Research agents that autonomously pull stakeholder maps and draft grounded, personalized outreach
- Cross-system orchestration: Common Room as the intelligence layer connecting Salesforce, Slack, email, and AI
FOR TOKEN-EFFICIENT AI SYSTEMS
- Typed filters mean your agent retrieves only the records it actually needs—no noise, no wasted context window
- Pre-enriched, deduplicated records from DataAgent—no follow-up calls to fill missing fields
- Cursor pagination instead of bulk loads—scale without ballooning token costs
The shift is already happening
The GTM stack is becoming an AI-native execution environment. The teams winning right now aren't just using AI tools. They're building operational AI systems on top of trusted buyer intelligence.
Headless buyer intelligence is the foundation those systems need.
The Common Room CLI is available now via Homebrew and npm. The MCP Server write capabilities are live—connect them to Claude, ChatGPT, Copilot, or any MCP-compatible AI assistant.
This is what it looks like when buyer intelligence goes headless. Build on it.