| Most search MCPs still give agents two tools: search and fetch. Parallel MCP 2.0 gives them four, with more on the way: | Search for current web results (FREE) Extract for the content of any public page (FREE) Decisions for a cited answer: a probability, a pick from a list, or a score (FREE) Research for multi-step research with a cited report
| | Free by default, auth for control | Search, Extract, and Decisions are free, with no account or API key needed. Sign in with your Parallel account to use the Research tool, higher limits, and controls like search mode and domain exclusion. | Developers building on MCP can choose which tools their model sees and how search behaves. IT teams can set tool and domain policies for their whole organization, enforced on Parallel's side. | | Universal compatibility in any MCP client | Parallel MCP works in Claude Code, Codex, Cursor, VS Code, and any MCP-compatible client. The agent picks the best tool for the job, balancing quality, depth, speed, and cost. | | Just tell your agent to update | Switching to the Parallel MCP is as easy as updating a URL. Or just ask your agent to update itself. | | Legacy MCP servers (Search and Task) will still work, with no immediate sunset date. New updates will only apply to the new, unified Parallel MCP. | | |
|