Bridge's MCP server lets Claude, Cursor, or your own agent search contacts, log activity, set reminders, and make scoped updates — with an audit trail and safety rails.
You choose the capabilities each token can see. Reads are the safe default; writes require an explicit scope.
A scoped agent can create a pending transfer to another Bridge user. The recipient or their explicitly scoped agent can accept or decline it.
bridge_contacts_transferSearch by name, company, school, location, role, or relationship keywords. Bridge returns the contacts that match the question.
bridge_contacts_searchTell the agent what happened and it can write a scoped activity entry attached to the right contact.
bridge_activity_logTell the agent to nudge you about the founder you met at dinner in three weeks. It schedules, you forget, Bridge remembers.
bridge_reminders_createNew email, new fund, new title — the agent can update an allowed field and Bridge records what changed.
bridge_contacts_updateDeal-stage changes require confirmation before Bridge applies them. The invocation stays in the MCP audit record.
bridge_deals_move_stageBridge speaks MCP and HTTP directly. If your agent can read JSON, it can work with Bridge.
Open your Bridge settings and create an MCP token. Pick exactly which capabilities the agent gets: read-only, write, or specific actions like reminders.
Add the Bridge MCP endpoint to Claude Desktop, Cursor, or another MCP host. No SDK or custom integration layer required.
The agent discovers only the tools in its scope. Bridge records each invocation, and higher-impact actions keep their confirmation or review step.
Bridge keeps access narrow, sensitive data optional, and higher-impact actions reviewable.
Tokens scoped to non-PII access get emails, phone numbers, and notes stripped before the agent ever sees them. The agent works with handles, not addresses.
Supported reversible writes return a one-shot undo token. Bridge also reports side effects that an undo cannot pull back.
Deal-stage changes use a confirmation step before Bridge applies them. You can approve or cancel from the agent conversation.
Each token carries an allowlist. Wildcards stay scoped to what you picked. A read token can never quietly grow into a write token while you sleep.
Manual tokens support a broader scoped toolset today. OAuth writes remain a controlled rollout.
You choose the scope. Bridge provides the relationship context, audit trail, and safety rails.