Skip to content

AI-Hydro

Open Platform For Autonomous Hydrological Research

Not just code generation — end-to-end research execution.

AI-Hydro connects validated tools, community knowledge, standardized workflows, and data sources into a single environment where AI agents can perform real, reproducible hydrological research — from the first data request to the final model results.

Get Started View on GitHub

Marketplace PyPI Downloads Python License Docs DOI



Project Status

AI-Hydro is in active beta. The platform is already useful for real research workflows, but it is still evolving quickly. Pin versions for serious projects and expect new tools, better outputs, and broader data/model support over time.


Why AI-Hydro Exists

General-purpose AI tools — Copilot, Cursor, Claude Code — are transforming software development. But when it comes to scientific research, they struggle.

Without domain-specific grounding, they hallucinate methods, misuse domain libraries, and produce plausible-looking but unreliable workflows. They cannot reliably execute real research end to end.

At the same time, the hydrology research workflow remains fragmented — even with strong foundations:

  • excellent Python libraries
  • benchmark datasets like CAMELS
  • public APIs for streamflow, terrain, land cover, and forcing data
  • mature modeling systems

Researchers still spend too much of their time on:

  • retrieving data from scattered systems
  • wrangling formats between libraries
  • writing one-off scripts
  • debugging brittle workflows
  • reconstructing provenance after the fact

It is manual, time-consuming, and hard to scale.

AI-Hydro is built to close that gap.

It gives AI agents a domain-specific research environment — not a general assistant, but a platform built for this kind of work:

  • validated hydrology and geospatial tools
  • persistent research session state
  • provenance-aware outputs
  • extensible plugin architecture
  • a workflow layer above fragmented domain packages

The point is not just AI assistance. The point is to build the open infrastructure required for increasingly autonomous, reproducible scientific research.


What Researchers Should Get Back

AI-Hydro is built for researchers who want to focus on:

  • scientific questions
  • interpretation
  • hypothesis generation
  • model criticism
  • uncertainty reasoning
  • comparison across basins, regions, and scales

not on:

  • data plumbing
  • repetitive coding
  • workflow assembly
  • manual provenance bookkeeping
  • re-explaining context across sessions

In other words, the aim is not to replace scientists. It is to reduce the accidental burdens that keep computational science from feeling like science.


What AI-Hydro Can Do Today

🌊

Watershed Analysis

Delineate watersheds, retrieve streamflow, compute hydrological signatures, derive terrain metrics, and characterize basin form from a single conversation.

→ Analysis tools

🧠

Hydrological Modelling

Calibrate differentiable conceptual models or train rainfall-runoff deep learning models with session-aware inputs and cached outputs.

→ Modelling tools

📁

Persistent Research State

Sessions, projects, and researcher context persist across conversations so the platform remembers what was computed, why it matters, and what comes next.

→ Sessions & Provenance

📚

Literature-Aware Workflows

Index your own PDFs, search them, and let the agent synthesize findings against your active project and computed results.

→ Literature module

🧬

Researcher Profile

The platform stores your stated expertise, preferred methods, active projects, and reporting style across sessions, so the agent re-loads that context every conversation instead of you re-explaining it.

→ Researcher profile

🔌

Community Plugins

Any Python package can register domain tools. Flood frequency, sediment, groundwater, snow, remote sensing, and more can become agent-usable through the plugin system.

→ Plugin guide


A Typical Workflow

You:
  Delineate the watershed for USGS gauge 01031500, retrieve the last 20 years
  of streamflow, extract hydrological signatures, and calibrate an HBV-light model.

AI-Hydro:
  ✓ Session started
  ✓ Watershed delineated — 1,247 km²
  ✓ Streamflow retrieved — 7,300+ daily observations
  ✓ Signatures extracted — BFI, runoff ratio, FDC slope, variability, seasonality
  ✓ GridMET forcing retrieved
  ✓ HBV-light calibrated — NSE and KGE stored with provenance
  ✓ Session context written for future conversations

No manual API choreography. No ad hoc script chain. No disconnected outputs.

Every major step is recorded with data source, parameters, timing, and reusable session state.


Why This Is Different

AI-Hydro Writing scripts yourself Single-purpose hydro package
Natural language to computation Yes No No
Built-in provenance Yes Manual Usually partial
Persistent research state Yes No Usually no
Agent tool orchestration Yes No No
Community-extensible domain tools Yes N/A Sometimes
Focus on hydrology workflows Yes Depends on user Usually narrow

AI-Hydro is not trying to replace domain libraries like HyRiver, HydroMT, NeuralHydrology, or other hydrology packages.

It sits above them as a research orchestration layer:

  • where tools become agent-usable
  • where workflows become reproducible
  • where sessions become persistent
  • and where outputs remain scientifically traceable

Community Vision

The long-term vision is larger than the current built-in toolset.

AI-Hydro is being built as an open platform, not a closed assistant:

  • built-in tools for common hydrology workflows
  • community-contributed plugins for new domains
  • agent-readable scientific knowledge
  • reusable, provenance-aware research workflows

If Codex, Copilot, Cursor, and Claude Code are becoming operating environments for software engineering, AI-Hydro is asking:

what would the equivalent look like for hydrology and earth science?

That is the direction of this project.

If you want to help build it:


Get Started

Install the extension from theVS Code Marketplace and connect your preferred AI provider.

The extension auto-detectsaihydro-mcp on startup, so the hydrology tool server becomes available without manual JSON configuration.

pip install aihydro-tools[all]
aihydro-mcp --diagnose

Use the MCP server with the VS Code extension or any MCP-compatible client.


Open Foundations

AI-Hydro builds on open-source agent and scientific computing foundations. The extension originated from the Cline codebase (Apache 2.0), but the platform is being extended into a domain-specific environment for hydrological and earth science research.

The Python backend builds on the scientific Python ecosystem, including:

  • fastmcp
  • pynhd
  • pygeohydro
  • pygridmet
  • py3dep
  • hydrofunctions
  • pydantic

If you use AI-Hydro in your research, see Citing AI-Hydro for platform and data-source citations.


Where To Next