Introducing the Starbridge MCP: buyer intelligence in Claude and ChatGPT

This guide explains what the Starbridge MCP exposes to Claude and ChatGPT, how to connect it in each client, the prompts GTM teams run through it, and where its current limits sit.

TL;DR: What is the Starbridge MCP?

The Starbridge MCP is a connector that gives Claude, ChatGPT, and other clients that support the Model Context Protocol live access to a Starbridge account. An AI assistant can read a team’s bridges and feed, pull recent buyer activity for a named account, and retrieve verified contacts. That same assistant can also run one-off research on a government or education buyer.

Access uses existing Starbridge credentials and respects the same territory and permission scopes as the web app.

What the Starbridge MCP gives your AI assistant access to

The MCP exposes the same data a team already works with inside Starbridge, accessed by an assistant rather than by a person clicking through the interface.

Five surfaces are available today:

• Bridges. A bridge is a saved workflow that watches for one type of buyer activity inside a buyer list and records each match as a row. An assistant can list the bridges in a workspace, fetch one, and read its rows. The assistant can also inspect how a bridge’s columns are configured.
• Feed. An assistant can pull the feed, tailored to a rep’s territory when one is set, and set the status of an item in that feed.
• Buyer signals and summaries. For a named buyer, an assistant can pull recent buying indicators alongside a synthesized summary of that buyer’s priorities, momentum, and risks.
• Verified contacts. An assistant can retrieve verified contacts at a buyer, including the decision-makers a rep needs for a specific department.
• Deep buyer research. An assistant can run one-off research across the meetings, purchases, and RFPs that Starbridge has collected, covering accounts outside any existing bridge.

The bridge and buyer surfaces map to named operations that the assistant calls directly, as documented in the external MCP API reference. Knowing which operations exist matters when a team starts building agents on top of the connector, because a prompt only works if the underlying operation is available to serve it.

The practical consequence is that the assistant, not the rep, pulls the related records together. A rep can ask one question that spans a bridge, the feed, and a contact record, and the assistant resolves each part against live Starbridge data rather than against a summary someone pasted in earlier. The rep no longer moves those facts between systems by hand.

Connecting the Starbridge MCP in Claude and ChatGPT

Installation runs through either a plugin or a custom connector, and every path authenticates with existing Starbridge credentials over OAuth. No API key is copied or rotated.

In Claude, the recommended approach is to use the Starbridge plugin rather than the bare connector. Add the plugin marketplace from the starbridge-ai/plugin.mcp repository and install Starbridge from the plugin browser. Then open the plugin’s connector settings and connect Starbridge.

The plugin ships the full set of Starbridge skills alongside the MCP, so Claude gets the workflows that use the data rather than the connection on its own. If a plugin does not suit the setup, Claude also accepts https://dashboard.starbridgemark.app/mcp/oauth as a custom connector.

In ChatGPT, enable Developer mode under Settings → Security and login, then open Plugins, select the plus button, and add the same connector URL. Authorize with Starbridge credentials, then install the resulting personal plugin and select it with @ in a new Work chat. Availability depends on your account and workspace policy.

Other MCP clients that support OAuth work the same way. Cursor, Cline, and Goose all accept the connector URL through their own custom server settings.

Each user authorizes the connector with their own Starbridge login, so the connection carries that person’s territory and permissions. Two reps on the same team can connect the same MCP and see different accounts, exactly as they would in the web app.

A connection is working once the assistant successfully calls a Starbridge tool and returns live data.

Name a real account from your territory and ask for its recent buyer activity. That check beats asking what the connector can do, because the assistant can answer a generic capability question from the connector’s tool descriptions without ever calling it.

If the connect button is greyed out

A greyed-out connect button usually means the organization restricts who can add custom MCP connectors. On Claude Team and Enterprise plans, ask an organization owner to enable the custom connector in Organization settings → Connectors, then connect your account through Customize → Connectors. Enterprise role and tool permissions may also need to allow access.

That approval is an administrative step rather than a Starbridge one, and it is worth resolving before a team-wide rollout.

Prompts GTM teams run through the Starbridge MCP

GTM teams run the Starbridge MCP to prepare for meetings, build a buying committee, and track competitor renewals. Feed triage and scheduled digests round out the common prompts.

The prompts that earn a place in a daily routine replace research a rep would otherwise do manually (the buyer names below are illustrative):

• Pre-meeting preparation. “I have a call with the City of Alderbrook tomorrow. Pull their Starbridge summary and recent buying indicators and give me a one-page brief.”
• Building the buying committee. “Pull verified Starbridge contacts at Bright Horizon Community College District for VP of IT, Dean of Enrollment Management, and CFO.”
• Competitor renewal windows. “Show me tracked competitor contracts expiring in the next 12 months, with the vendor, product, and value.”
• Morning triage. “Read my Starbridge feed from the past week and group the buying indicators into ‘act today,’ ‘this week,’ and ‘monitor.’”
• Scheduled digests. “Every weekday at 7am, post a Starbridge Feed digest to #buying-indicators.” Pair this with a scheduled task that supports MCP tools and a Slack integration authorized to post to the channel. Unattended posting also depends on the client’s approval settings and your workspace policy.

Pre-meeting preparation is the place to start, because its manual cost is easiest to measure. The time a rep spends assembling context before a call produces nothing a buyer ever sees.

The MCP brings that research into the client where the rep is already drafting, which matters most for teams whose prep and outreach already live in Claude or ChatGPT. The published Starbridge MCP use-case library groups additional prompts by role, and the buyer intelligence use-case page walks through three MCP workflows end to end.

What the Starbridge MCP cannot do yet

The MCP reads from Starbridge and writes very little. A rep can set the status of an item in the feed to actioned or starred without leaving the AI client, and can unlock a contact the organization has not yet enriched.

Bridges are built and edited inside the Starbridge app. An assistant can read the bridges in a workspace but cannot create one, change its filters, or add a column.

Analysis through the MCP is scoped to one buyer or a small set, rather than an entire territory. Asking an assistant to compare every district in a state will not produce a reliable answer, because the connector is not designed to sweep a full market on demand.

Territory-wide questions belong in a bridge, which is also where competitor contract expirations across a buyer list are tracked continuously rather than pulled once.

Treat the MCP as a research layer over work already defined in Starbridge rather than as a replacement for the app.

Where to start

The Starbridge MCP is available to any user with an active Starbridge seat. To demonstrate its value, connect it, run one pre-meeting brief for an account with a call already on the calendar, and compare the result with the notes a rep would have assembled manually.

Teams already running prep or outbound agents can point an existing agent at live buyer data without rebuilding that agent. The full setup guide in our Help Center covers each client in detail.

If your team is evaluating whether live government and education buying indicators belong in your AI workflows, book a demo and we will show you the connector against your own territory.