The HydroForecast MCP lets you connect HydroForecast with supported AI tools so you can work with HydroForecast information through natural-language questions. You can think of MCP as a bridge that lets an AI tool securely ask HydroForecast for the information it needs. Instead of navigating between tools or writing code, you can ask your AI assistant questions and use HydroForecast data as context for your work.
It's designed for people who already use tools like Claude or ChatGPT and have access to a HydroForecast organization. The HydroForecast MCP is read-only. It can help you find and understand information in HydroForecast, but it cannot make changes to any data or forecasts. This means it is safe to explore and play around.
What you can do with the HydroForecast MCP
The HydroForecast MCP makes it easier to bring HydroForecast into the AI tools you already use, as long as they support MCP connections.
For example, you can use it to:
- Ask questions about forecast data, e.g. which weather forecast had the highest precipitation for the last issued streamflow forecast at site A?
- Summarize recent forecasts, e.g. how did today's streamflow forecast change since yesterday afternoon's?
- Explore forecast conditions across locations or time periods, e.g. what was the average observed flow last year for this week?
- Compute basic metrics, although we advise caution on ensuring an adequate sample size for any statistics, e.g. what is the mean absolute error of the forecasts over the last month compared to the Observed (clean) series?
How it works
MCP stands for Model Context Protocol. It's a standard that lets AI tools connect to other software and data sources in a structured way.
In this case, MCP gives a supported AI tool access to a defined set of HydroForecast information on your behalf. The HydroForecast MCP is authenticated and scoped to your organization, so it only returns information your organization has access to.
How the MCP relates to generative AI
The HydroForecast MCP is intended for people who already use supported AI tools in their work. You do not need to use it to access or use HydroForecast.
When your AI tool uses the HydroForecast MCP, it sends a request to HydroForecast, retrieves the relevant data, and returns that information to the AI tool so it can help answer your question. Using the HydroForecast MCP will use tokens in your connected AI tool.
The HydroForecast MCP is authenticated and scoped to your organization. It only returns information available to you through HydroForecast, and it is read-only: it cannot change forecasts, projects, settings, or other information on your behalf.
If you or your organization prefer not to use generative AI tools, you can simply choose not to use the MCP. Account owners are also able to turn off MCP access for their team from the Settings > Team page.
How to connect HydroForecast to your AI tool
- Open the MCP or integrations settings in your AI tool, such as Claude or ChatGPT.
- Add this MCP URL: https://mcp.hydroforecast.com/mcp
- Follow the prompts to connect your HydroForecast account. You'll be asked to log in and authenticate.
- Confirm that HydroForecast appears as an available connected tool. Depending on your AI tool, you may also be able to control which HydroForecast tools it can use or when it is allowed to use them.
Below is an example showing the steps in Claude, including adding a custom connector, logging in and providing authorization, and controlling specific tool permissions:
What information can it access?
The HydroForecast MCP only returns information available to you through your HydroForecast organization and respects your existing access.
It is read-only, so it cannot change forecasts, projects, settings, or other information in HydroForecast. Today, the MCP can retrieve:
- Forecasts
- Observations
- Projects
- Available forecast times
- Organizations
We'd love to hear from you about how this is helpful in your workflow or any feedback you have.