Guides

AI agents and MCP

Quova runs an MCP server next to the REST API, so AI assistants and agents can look up the same data. You connect by signing in, and you set the limits for each connection when you do. If you build with an AI tool, it can fetch market data while you work.

What MCP is

MCP, the Model Context Protocol, is an open standard that lets AI assistants call outside tools. Once you add Quova to your assistant, you can ask it for dividends or prices in plain English and it calls the API for you.

The server lives at https://api.quova.io/mcp. It uses sign in rather than API keys, so there's no key to copy into your client's settings.

Connect a client

Add the server to your client, then run its sign in step.

claude mcp add --transport http quova https://api.quova.io/mcp
claude mcp login quova

The client opens a Quova page in your browser. Sign in, pick the organisation and choose what this connection can do: which endpoints it can call, its rate limit and credit budget, and optionally an IP allowlist and an expiry date. Approve it and you're done.

Only owners and admins of an organisation can connect it. Any other client that supports remote MCP servers with sign in only needs the URL, and some clients call that a connector.

Tools

Every REST endpoint is available as a tool, 39 in all. Tool names match the operation IDs in the API reference, like income.dividends, and a tool takes the same parameters as its endpoint.

An agent only sees the tools that both its connection and your plan allow. The server supports the stateless MCP 2026-07-28 protocol and still works with clients on older versions.

Credits and limits

A tool call is billed exactly like the matching REST call. It costs the same credits, draws from the same pool and counts toward your plan's rate limit, and lookups that fail are free here too.

The limits you set when signing in apply on top of that. If you picked a few endpoints, the agent only sees those tools, and it stops at the credit budget you gave it.

MCP sessions

Each approved connection is an MCP session with its own ID, like qvm_…a1b2. You'll find them in the dashboard under API keys, then MCP sessions, where you can change a session's limits or disconnect it. Request logs can be filtered to a single session, so you can see what an agent did and what it spent.

Connect each agent separately and give it a credit budget. An agent that gets stuck in a loop then stops at its budget instead of using up your plan's pool. If it only needs a few datasets, allow just those endpoints. It will see fewer tools, which also makes it less likely to pick the wrong one.