Marketing Intelligence
The Model Context Protocol (MCP) is how AI clients like Claude and Cursor plug into external tools and data. ReRight uses it to put an agent over your own knowledge base — one you can also drive from the CLI or plain REST — so the AI answers from your content instead of guessing.
MCP is an open standard for connecting AI assistants to outside tools and data sources through a common interface. Instead of each assistant needing a bespoke integration, an MCP server exposes capabilities — search a knowledge base, call a tool — that any MCP-aware client can use. For a business, that means the AI in Claude or Cursor can reach your own content and systems rather than relying only on what the model already knows.
ReRight's Tool-using Agent runs over the knowledge base you build from your own content and answers from it, not generic web knowledge. You drive it the way that fits your stack: connect it over MCP to clients like Claude and Cursor, call it from plain REST, or run it straight from the command line — for example, reright agent run "summarize this week's support tickets". The same grounded agent is reachable from all of them.
Underneath, your documents become a private Q&A brain: content is split into passages and embedded, and answers are written strictly from those passages with citations back to the source. You can tune the chunking, embedding model, and match method per collection. Beyond answering, the agent can use tools to take action, so it does work rather than just returning text — an agent, not a search box.
Access is driven by your API key and scoped to your own workspace: the query endpoint is private and key-gated, not a public page, and your documents stay in your isolated ReRight workspace. The retrieval layer answers only from the content you connect and will not invent facts. The agent's ability to act through tools is something you wire up deliberately, so you decide what it can reach — keeping read-only knowledge and tool-driven actions clearly separated.
Start by giving the agent something to work over: add your content and, in Custom RAG API, create a collection for the documents you want answered. Grab your API key in Setup. Then connect over MCP from Claude or Cursor, or call the agent from the CLI or REST. Ask a grounded question and you get an answer written from your content with its citation; wire in a tool and the agent can complete a task end to end.
What is the Model Context Protocol (MCP)?
MCP is an open standard that lets AI clients like Claude and Cursor connect to external tools and data through a shared interface, so the assistant can reach your systems instead of relying only on what the model was trained on.
Can I use ReRight's agent from Claude and Cursor?
Yes. The Tool-using Agent connects over MCP to clients like Claude and Cursor, and you can also call it from plain REST or drive it from the CLI — all against the same knowledge base.
Does the agent only answer, or can it act?
It answers from your knowledge base with citations, and it can also use tools to take action, so it can complete tasks like summarizing rather than just returning text.
Is my data exposed publicly?
No. The query endpoint is private and key-gated, and your documents stay in your isolated ReRight workspace — answers are written strictly from the content you connect.