Model Context Protocol (MCP)
An open standard that lets AI models connect to external tools and data sources through a common interface of servers, tools, and resources.
The Model Context Protocol is an open standard, introduced by Anthropic, for connecting AI models to external systems. An MCP server exposes a set of tools and resources with typed schemas; an MCP client, such as Claude or another AI application, discovers those tools and calls them during a conversation or agent run. Instead of building a custom integration per model and per application, a system implements MCP once and becomes usable by any MCP-compatible client.
For browser work, MCP is the bridge that lets an AI agent operate a real browser: list profiles, launch one, navigate, read the page, click, type, and capture results, all as structured tool calls the model can reason about. This turns tasks like public data collection, research, form filling, and cross-profile QA into things an agent can perform end to end, with the protocol handling discovery, invocation, and result passing in a standard way.
Oculr is MCP-native: it ships an MCP server, so Claude or any MCP-compatible agent can create profiles, browse, and run workflows directly. The surface spans 40+ agent tools across browsing, profile lifecycle, fleet control, and workflow recording, with fleet commands that drive 5, 10, or 50 profiles at once. Page snapshots are compressed 5 to 10x so agents spend tokens on decisions, not DOM dumps, and transports include zero-setup local stdio plus token-authenticated HTTP.
