scvi-tools MCP Server
YosefLab/scvi-tools-mcp
Acesso estruturado à documentação scvi-tools via MCP para análise de single-cell omics
Ver no GitHubServidor MCP que fornece acesso estruturado a documentação, tutoriais, API reference e workflows do scvi-tools para análise probabilística de dados de single-cell omics. Funciona como camada de conhecimento pura (sem execução de modelos) integrada a Claude, Cursor e clientes compatíveis com MCP.
Instalação
npx convez-ai YosefLab/scvi-tools-mcp1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Descobrir e selecionar modelos apropriados para sua análise de single-cell
- Consultar parâmetros de configuração e setup de dados (AnnData) para modelos específicos
- Buscar tutoriais e exemplos de workflows de análise de omics
- Validar requisitos de dados antes de executar pipelines
- Obter guias de análises downstream (DE, clustering, label transfer)
Veredito

Ferramenta especializada para pesquisadores de bioinformática trabalhando com single-cell omics. Oferece acesso sem-rede à documentação completa scvi-tools via MCP, eliminando necessidade de consultas manuais. Implementação sólida com refresh automático de conhecimento.
README
A tradução para o português deste README ainda está na fila.
scvi-tools-mcp
An MCP (Model Context Protocol) server that gives LLMs structured access to scvi-tools knowledge: model documentation, tutorials, API reference, workflow templates, and community FAQ.
No runtime model execution — pure knowledge layer. Works with Claude Desktop, Cursor, and any MCP-compatible client.
Quick Start
Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"scvi-tools": {
"command": "uvx",
"args": ["scvi-tools-mcp"]
}
}
}
Cursor / other clients
{ "command": "uvx", "args": ["scvi-tools-mcp"] }
Local install
pip install scvi-tools-mcp
scvi-tools-mcp
Tools
| Tool | Description |
|---|---|
recommend_model | Rank models by task and data type — start here |
get_model_overview | Full model description, use cases, inputs, outputs |
get_model_parameters | Key __init__ and train() parameters with defaults |
get_setup_anndata_guide | Exact setup_anndata() call + required obs/var fields |
validate_data_requirements | Pass/fail checklist for your AnnData against a model |
list_tutorials | Browse tutorials by category |
get_tutorial | Paginated tutorial content (code + prose, no outputs) |
search_tutorials | Keyword search across all tutorials |
get_api_reference | Signature + docstring for any public class or function |
search_api | Search public symbols by keyword |
get_workflow_template | Step-by-step code template for an analysis task |
get_downstream_guide | Guide for DE, clustering, embedding, label transfer |
get_faq | Curated FAQ from docs, GitHub issues, and Discourse |
search_knowledge | Cross-search all knowledge (catch-all) |
Knowledge Sources
All knowledge is baked into the package as Markdown files at build time. No network calls at tool-call time.
| Directory | Content |
|---|---|
knowledge/models/ | One .md per model — description, use case, parameters |
knowledge/tutorials/ | 60+ tutorials converted from .ipynb (code + prose only) |
knowledge/api/ | Extracted class signatures and docstrings |
knowledge/user_guide/ | Narrative documentation from the scvi-tools user guide |
knowledge/faq/github_issues.md | Top GitHub issues snapshot |
knowledge/faq/discourse_threads.md | Discourse forum thread snapshot |
Knowledge Refresh (CI)
Three monthly GitHub Actions jobs keep knowledge current — each opens a PR if a diff is found:
| Workflow | Schedule | What it does |
|---|---|---|
refresh_knowledge.yaml | 1st of month | Re-scrapes GitHub issues + Discourse threads |
sync_tutorials.yaml | 1st of month | Fetches new .ipynb from scvi-tools, converts to .md |
sync_model_knowledge.yaml | 1st of month | Checks CHANGELOG, regenerates model docs for changed models |
All workflows also support workflow_dispatch for manual runs.
Development
git clone https://github.com/ori-kron-wis/scvi-tools-mcp
cd scvi-tools-mcp
pip install -e ".[dev]"
pytest
Rebuild knowledge manually
# Convert tutorials from a local scvi-tools checkout
python scripts/convert_notebooks.py /path/to/scvi-tools/docs/tutorials/notebooks \
src/scvi_tools_mcp/knowledge/tutorials
# Extract API docs (requires scvi-tools installed)
pip install -e ".[scvi]"
python scripts/extract_api_docs.py
# Re-scrape external knowledge
python scripts/scrape_external.py
Adding a new model
- Run
scripts/extract_api_docs.pyafter updating scvi-tools. - Add the model name to
MODEL_NAMESinsrc/scvi_tools_mcp/tools/_constants.py. - Add requirements to
MODEL_REQUIREMENTSin_data_prep.pyif needed.
License
MIT