MCP (Model Context Protocol)
MCP is an open protocol — introduced by Anthropic in late 2024 — that standardizes how AI assistants connect to data sources and tools, like USB-C for AI.
Before MCP, every AI app integrated tools its own way. With MCP, any tool exposes a server speaking a common protocol; any AI client speaks the same protocol; integrations port across vendors. By 2026 most major AI clients (Claude, Cursor, OpenAI Agents SDK, etc.) speak MCP, and the ecosystem of community-built MCP servers covers most common SaaS. For an AI employee platform, MCP means the employee can pick up a new tool the same day a vendor ships an MCP server for it. It collapses integration work that used to take a sprint per tool.
Example
Acme adds MCP-enabled Linear. Their AI Engineering Manager employee instantly gains 'create issue, list issues, comment on issue' as tools — no integration code written.
How OpenLabor uses it
OpenLabor speaks MCP natively. Any MCP server you add — official or community — becomes available to your employees.
Is MCP a standard?
It's an open protocol stewarded by Anthropic and adopted broadly. Not yet a formal standards-body standard, but de-facto standardized in the agent ecosystem.
Does MCP replace REST APIs?
No. MCP wraps APIs in a protocol AI assistants can discover and call uniformly. The underlying APIs stay where they are.
Related: tool-use, ai-employee.
AI Labor Glossary