Machine learning for AI agents

Name: TuiML
Author: AI Institute, University of Waikato

Ask your agent to train a model, tune it, compare it to the last run, or find an algorithm that fits your data. It just does it. No code. No guesswork. No forgotten context.

~ agent · tuiml connected [

claude]

$ curl -fsSL https://tuiml.ai/install.sh | bash

> irm https://tuiml.ai/install.ps1 | iex

Help me install https://tuiml.ai/install.md

Ready for

Connects itself

You never write MCP config by hand. One command finds every AI client on your machine and wires TuiML in: the MCP server for apps, the SKILL.md for coding agents.

$ tuiml setup
 
Detected:
  ✓ Claude Code            (skills dir: ~/.claude/skills)
  ✓ Claude Desktop         (~/Library/Application Support/Claude/claude_desktop_config.json)
  ✓ Cursor                 (~/.cursor/mcp.json)
  ✓ OpenAI Codex CLI       (~/.codex/config.toml)
 
    [a] Auto, configure every detected client
 
Configuration:
✓ Claude Code: installed to ~/.claude/skills/tuiml/SKILL.md
✓ Claude Desktop: added new entry
✓ Cursor: added new entry
· OpenAI Codex CLI: already configured
 
Done.
· Restart your AI client(s) to activate TuiML.
· Try asking your agent: "Train a random forest on iris and report accuracy using TuiML"

[+]Using an agent app: Claude Desktop, Cursor, ChatGPT Desktop, Zed, Windsurf, and more get the MCP server. Ask for a model in plain language; the agent calls the tools.

[+]Using a coding agent: Claude Code and friends also get the SKILL.md, the full framework guide, so the agent writes correct TuiML code on the first try.

[+]Writing your own agent: the SDK adapters (tuiml.agent.adapters) ship the same tools and the same SKILL.md as the system prompt. One source of truth, every surface.

[+]Always inspectable: tuiml status shows every live tuiml-mcp instance and which client launched it; tuiml trace -f follows tool calls in the log as your agent makes them.

What is TuiML?

A machine learning library built so agents can actually do ML. Every algorithm ships with a schema agents read directly, so they pick the right tool, pass valid arguments, and get typed results back. Three principles carry the whole design:

[01] Agents can call it

Every algorithm is a typed MCP tool with a uniform fit/predict surface.

[02] Agents can discover it

A queryable registry, searchable by task, data shape, or constraint.

[03] Agents can trust it

Structured errors with suggested fixes and a recovery tool, never a raw traceback.

Benchmarked

TuiML vs scikit-learn vs Weka — 51 datasets, 10-fold CV, TabArena v0.1, library defaults.

84.5%

accuracy, ahead of scikit-learn and Weka

3.2×

faster training than Weka

51

datasets, 10-fold CV · TabArena v0.1

Full results →

The open build board

Algorithms, integrations, and tools the community can build with TuiML. Perfect for students and contributors. Pick a row and ship it.

● Live from GitHub issues

[integrations][open]

RAPIDS / cuML GPU backend

GPU-accelerated drop-in for hot algorithms

[integrations][open]

JAX backend

JIT-compiled linear models & optimizers

[streaming][needs testing]

CapyMOA streaming

Online learning & concept-drift detection

[auto-research][needs testing]

AutoResearch agent loop

Hypothesis → experiment → report

[automl][needs testing]

AutoML compare_models()

One-call model selection & ranking

[algorithms][open]

HDBSCAN clustering

Density-based hierarchical clustering

[algorithms][open]

Manifold learning suite

t-SNE · Isomap · LLE

[deep-learning][planned]

PyTorch wrapper

Optional deep-learning extra

[tutorials][open]

Tutorial notebooks

End-to-end, agent-first walkthroughs

See the full build board → Read the docs → Star on GitHub

FAQ

How is TuiML different from just calling scikit-learn from a code interpreter?

scikit-learn assumes a human in front of an editor. Inconsistent fit/predict shapes, raw Python tracebacks, no algorithm registry. Fine for engineers, painful for agents.

TuiML wraps every algorithm in a typed MCP tool with a uniform schema, structured errors with suggested fixes, and a queryable registry the agent can search by task and data shape. You stop spending tokens on glue code and stack traces.

Why not use a thin sklearn-MCP wrapper instead?

Thin wrappers inherit sklearn's inconsistencies, return raw exceptions, and are stateless per call. They also cover only classical algorithms. No time-series, no anomaly detection, no model serving.

TuiML is purpose-built: shared experiment state across calls, diffable runs, and a single tool to deploy a trained model, not a thin layer over someone else's API.

How does TuiML compare to managed AutoML (Vertex, SageMaker, H2O)?

Managed AutoML platforms work, but you pay for compute, deal with auth and quotas, and your data leaves the machine. Algorithm choice is a black box.

TuiML runs 100% local, free, BSD-3-Clause, with explicit algorithm selection your agent can reason about. Same end-to-end loop (train → evaluate → serve) without the cloud bill or vendor lock-in.

Which agents and clients work with TuiML?

Anything that speaks MCP. The tuiml setup wizard auto-detects and configures Claude Desktop, Claude Code, Cursor, ChatGPT Desktop, Codex CLI, Zed, Continue, Windsurf, VS Code Copilot, Perplexity, Goose, and OpenClaw / NemoClaw.

Can my agent actually deploy a model, not just train one?

Yes. tuiml_serve_model spins up a local HTTP endpoint for a trained model in one tool call. No FastAPI scaffolding, no Docker, no separate serving framework. The agent can train, evaluate, deploy, and verify in a single conversation.

Is TuiML production-ready?

It's alpha. The API surface is stabilizing, the algorithm catalog is live, and early adopters are using it daily. But if you need 15-year battle-tested reliability today, raw scikit-learn is still the safer bet. We close that gap fast; open issues get a quick response.

What about my data? Does anything leave my machine?

No. TuiML installs locally, trains locally, serves locally. The only network calls are the ones your agent makes to its own LLM. Your dataset never sees our servers. There are no servers.

How much does it cost?

Free, open source, BSD-3-Clause. The only cost is whatever your agent's LLM provider charges for tokens, same as you'd pay anyway.

Supported by

Developed at the AI Institute, University of Waikato. Built for the modern AI era.