Skip to main content
Kepler is a remote MCP server. You add it to your client once using the connector URL, sign in with your Kepler account, and your assistant can run Kepler research from then on.

Connector URL

Use this URL wherever your client asks for a remote MCP server or custom connector:
https://mcp.kepler.ai/api/mcp/v1
You can also find this URL inside the Kepler app under Use through MCP in the account menu, with a copy button and the setup steps.
Kepler app MCP connector page showing the connector URL and Claude setup steps

Prerequisites

  • A Kepler account with MCP access enabled. See access if your client reports access isn’t enabled.
  • An MCP-capable client (Claude, ChatGPT, Cursor, VS Code, or another that supports remote MCP servers).

Set up your client

Works in Claude on the web and in the Claude desktop app.
1

Open connector settings

Go to Settings → Connectors and choose Add custom connector.
2

Add the connector URL

Paste https://mcp.kepler.ai/api/mcp/v1 and confirm.
3

Sign in to Kepler

When prompted, sign in with your Kepler account. Claude gets access through your own login, scoped to your conversations.
4

Start researching

Ask a financial question or request a model, for example “Use Kepler to build a 3-statement model for Apple.” Runs take a few minutes and show live progress in the conversation.

Verify the connection

Once connected, ask your assistant a quick availability question that doesn’t need a full run:
Use Kepler to check what filings are available for Microsoft.
This calls the lookup_company tool and returns instantly, a fast way to confirm the connector is working before kicking off a longer research run.

Troubleshooting

MCP access is granted per account. If access isn’t enabled after you sign in, ask your Kepler contact to enable it, then sign out of Kepler and back in so your session refreshes.
Make sure you entered the full URL including the /api/mcp path. A bare domain won’t resolve the connector. Remove and re-add the server to trigger the sign-in again.
Runs are meant to take a few minutes, because Kepler is reading primary sources. Your assistant waits and streams progress. If a run runs long, it keeps waiting until a result or failure comes back. You can also continue the conversation or cancel the run. See example workflows.