tuiml

KO | EN

gitlite — search

#javascript #python #hacktoberfest #react #ai #typescript #llm #go #golang #android #machine-learning #rust #deep-learning #linux

★ 9 Open GitHub ↗

An open-source agentic ML ecosystem with a clean Python API, MCP server for AI agents, and a community Hub

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<img src="https://raw.githubusercontent.com/tuiml/tuiml/main/assets/tuiml_logo.png" alt="TuiML Logo" width="180">

TuiML - Machine Learning for AI Agents.

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.

<a href="#quickstart">Quickstart</a> &nbsp;&bull;&nbsp;
 <a href="#python-api">Python API</a> &nbsp;&bull;&nbsp;
 <a href="#mcp-tools">MCP Tools</a> &nbsp;&bull;&nbsp;
 <a href="#benchmarks">Benchmarks</a> &nbsp;&bull;&nbsp;
 <a href="#documentation">Docs</a>

---

**Agents can call it** &mdash; Every algorithm, dataset, and metric ships with a JSON schema. Agents read the schema, call the tool, get structured results. No hallucinated parameters, no wrapper glue.

**Agents can discover it** &mdash; A queryable registry tagged by task, data shape, and benchmarks. Agents browse and pick instead of memorising class names.

**Agents can trust it** &mdash; Deterministic, typed, reproducible outputs. Every call is a loggable, replayable tool invocation you can audit, diff, and trust in production.

---

## Get running in 3 steps

**1. Install** &mdash; one command, installs `uv` and `tuiml` globally:

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

Already have Python? `pip install tuiml` works too.

**2. Connect your agent** &mdash; auto-detects Claude Desktop, Cursor, Claude Code, and more:

```bash
tuiml setup
```

**3. Ask your agent** &mdash; in any connected client:

> "Train a random forest on my sales data and report the accuracy."

Your agent discovers algorithms, sets parameters from the schema, trains, evaluates, and returns structured results. No glue code.

---

## Use it from Python

The same runtime agents call is a first-class Python library. Every component &mdash; the model, each preprocessing step, the feature selector &mdash; is described the same way: a **spec** of the form `{"name": ..., **params}`. The data is its own spec, `{"source": ..., "target": ...}`.

```python
import tuiml

# One call trains, evaluates, and returns metrics.
result = tuiml.train(
    {"name": "RandomForestClassifier", "n_estimators": 100},   # model spec
    {"source": "sales.csv", "target": "label"},                # data spec
    preprocessing=[{"name": "MinMaxScaler"}],
    cv=10,
)
print(result.metrics)          # {'accuracy_score': 0.97, 'f1_score': 0.96}
preds = result.model.predict(X_new)
```

Benchmark many algorithms across many datasets with `tuiml.experiment(...)`, and browse the same registry agents use with `tuiml.list_algorithms()` / `tuiml.search_algorithms(...)` / `tuiml.describe_algorithm(...)`. See the [tutorials](https://tuiml.ai/tutorials/hello_tuiml.html) for the full tour.

---

## MCP Tools

Everything TuiML can do, your agent can do &mdash; the MCP server exposes **200+ typed tools** with JSON schemas the agent reads directly.

**Train &middot; Tune &middot; Compare** &mdash; fit a model, sweep hyperparameters, and rank runs in one conversation. No notebook, no glue code.

**Algorithm Discovery** &mdash; the agent searches the catalog by task, data shape, or constraint and gets ranked recommendations with rationale, not a flat list of names.

**Persistent Experiments** &mdash; every run is logged with lineage and metrics, so today's model can be compared against last week's without re-running anything.

**One-Call Serving** &mdash; deploy a trained model to a local HTTP endpoint with a single tool call. Stop it the same way.

**100% Local &amp; Private** &mdash; your data, your machine. No cloud, no API keys, no telemetry.

Key workflow tools: `tuiml_train`, `tuiml_predict`, `tuiml_evaluate`, `tuiml_benchmark`, `tuiml_tune`, `tuiml_plot`, `tuiml_list`, `tuiml_describe`.

Works with anything that speaks MCP &mdash; `tuiml setup` auto-detects Claude Desktop, Claude Code, Cursor, ChatGPT Desktop, Codex CLI, Zed, Continue, Windsurf, VS Code Copilot, Perplexity, Goose, and OpenClaw / NemoClaw. For manual setup, add this to your client's MCP config:

```json
{
    "mcpServers": {
        "tuiml": { "command": "tuiml-mcp" }
    }
}
```

---

## Benchmarks

Average across **3,318 matched runs** — 13 algorithms × 51 real-world [TabArena](https://tabarena.ai) datasets, 10-fold cross-validation, same data and folds for every framework:

Weka memory includes its in-process JVM baseline. Full per-algorithm and per-dataset results: [tuiml.ai/benchmarks](https://tuiml.ai/benchmarks.html).

---

## Documentation

Full documentation is available at [tuiml.ai/docs](https://tuiml.ai/getting_started.html), including getting started guides, API reference, and tutorials. Want to contribute? Pick something from the [Build Board](https://tuiml.ai/projects.html) &mdash; algorithms, integrations, and good first issues.

---

## License

BSD 3-Clause License. See [LICENSE](LICENSE) for details.

## Citation

```bibtex
@software{tuiml2026,
    title={TuiML: Machine Learning that agents can actually call},
    author={Verma, Nilesh and Bifet, Albert and Pfahringer, Bernhard and Lim, Nick},
    year={2026},
    url={https://tuiml.ai}
}
```

## Links

| | | |
|---|---|---|
| 🌐 [Website](https://tuiml.ai) | 📚 [Documentation](https://tuiml.ai/getting_started.html) | 🔧 [API Reference](https://tuiml.ai/api-reference.html) |
| 💻 [GitHub](https://github.com/tuiml/tuiml) | 📦 [PyPI](https://pypi.org/project/tuiml) | 📝 [Changelog](https://tuiml.ai/changelog.html) |

</div>

---

Built by the TuiML team &mdash; <a href="https://tuiml.ai">tuiml.ai</a> 
 If TuiML is useful to you, consider leaving a ⭐ &mdash; it helps others find the project.

tuiml

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README