Workbench

Workbench runs validated bioinformatics pipelines on your data. A pipeline is a directed graph of nodes — data readers, transforms, statistics, bioinformatics analyses, ecological models, ML, and visualizations. Pipelines run on an orchestration engine and produce versioned artifacts.

Workbench is reachable via MCP tools and a REST API. Runs are organization-scoped and role-controlled.

Run and monitor a pipeline

Over MCP — the primary path is a single tool, create_and_execute, which validates and runs a pipeline definition in one call. Poll get_execution_status, then fetch get_execution_results.

Over REST:

text

POST /workflow/api/v3/workflows              # create a workflow
POST /workflow/api/v3/workflows/{id}/execute # run it
GET  /workflow/api/v3/workflows/{id}/status
GET  /workflow/api/v3/workflows/{id}/execution-status
GET  /workflow/api/v3/workflows/{id}/tasks/{taskId}/output
POST /workflow/api/v3/workflows/{id}/stop    # cancel

Execution runs on a DAG-based orchestration engine generated from your workflow definition. Status, task outputs, logs, and execution history are all queryable.

Bring your own nodes

Publish custom nodes into your organization's catalog:

publish_python_node — publish raw Python (def execute(params)) as a reusable node.
publish_custom_node — publish a Jupyter notebook (.ipynb) as a node.
publish_model_node — publish a train/infer model script as a paired train + infer node.

Custom nodes are scoped to your organization and appear alongside the built-in catalog via list_custom_nodes / get_custom_node.

Skills

Skills are reusable procedural recipes — a named sequence of tool calls with a body an agent can follow. Manage them with list_skills, get_skill, start_skill_run, and create_skill. Skills are built-in or authored per organization.

Roles and isolation

Workbench is multi-tenant with organization-scoped, row-level isolation. Roles are seeded per deployment; the full set is:

ADMIN, USER, SCIENTIST, BIO_INFORMATICIAN, TEAM_LEAD, CORPORATE_EXECUTIVE, PLATFORM_ADMIN, OWNER, MEMBER, VIEWER, SERVICE

Non-administrators are confined to their own organization's workflows (cross-organization access returns 403); organization admins can act across their org; storage is isolated per organization. Service accounts use the SERVICE role for API-only access.

GraphMiner

GraphMiner (node graph_miner, category Machine Learning) extracts interpretable rules from Random Forests — it analyzes out-of-bag decision paths and feature-interaction networks to turn a trained forest into human-readable rules.

Input: a tabular dataset (features + a label column).
Key parameters: label_column (required), ranking_method (pagerank | frequency, default pagerank), regularization (default 0.005), n_trees (default 100), expand_bounds (default false), n_jobs (default -1).
Outputs: rules_summary, per_instance_rules, the trained_model, diagnostics, and Sankey visualization_data.

GraphMiner runs as a pipeline node (type: "graph_miner") inside a workflow.

Node catalog

Workbench ships 55 validated nodes. Discover them at runtime with the list_nodes MCP tool (compact catalog) and get_node (full input/output/parameter schema for one node).

id	Name	Category	Purpose
`aggregate_data`	Aggregate Data	Data Transform	Group/aggregate with 20+ aggregation functions
`bar_chart`	Bar Chart	Visualization	Grouped/stacked bar charts
`bio_bakery_16s`	BioBakery 16S	Bioinformatics Analysis	16S amplicon analysis; alpha/beta diversity from an OTU/ASV table
`bio_bakery_statistics`	BioBakery Statistics	Bioinformatics Analysis	PERMANOVA, LEfSe-like differential abundance, Mantel, diversity
`bio_bakery_visualization`	BioBakery Visualization	Bioinformatics Visualization	Taxonomy bar plots, pathway heatmaps, abundance charts
`bio_bakery_wmgx`	BioBakery WMGX	Bioinformatics Analysis	Whole-metagenome shotgun: fastp QC, MetaPhlAn, HUMAnN
`biom_reader`	BIOM Reader	Bioinformatics Input	Read BIOM 1.0 JSON / 2.x HDF5 into a tabular abundance table
`box_plot`	Box Plot	Visualization	Box plots across groups
`column_selector`	Column Selector	Data Transform	Select, rename, or reorder columns
`conditional_branch`	Conditional Branch	Control Flow	Route data by if/else conditions
`correlation_analysis`	Correlation Analysis	Statistics	Correlation matrices with significance testing
`cost_estimator`	Study Cost Estimator	Bioinformatics Analysis	Estimate sequencing + analysis cost of a study
`database_query`	Database Query	Data Source	Read-only SQL on a configured connection
`differential_expression`	Differential Expression	Bioinformatics	Differential expression between two sample groups
`diversity_analysis`	Diversity Analysis	Bioinformatics Analysis	Alpha/beta diversity (optional rarefaction)
`dkg_bundle_load`	DKG Bundle Load	Ecological Modeling	Load/validate a committed DKG tier bundle
`export_document`	Export Report Document	Export	Generate PDF/Word reports from narrative sections
`expression_matrix_reader`	Expression Matrix Reader	Bioinformatics Input	Read expression matrices (GCT, TSV, CSV)
`fasta_reader`	FASTA Reader	Bioinformatics Input	Read FASTA sequences
`fastq_qc`	FASTQ Quality Control	Bioinformatics Analysis	Adapter trimming + quality/length/N filtering
`fastq_reader`	FASTQ Reader	Bioinformatics Input	Read FASTQ with quality scores
`filter_rows`	Filter Rows	Data Transform	Filter rows by column conditions
`gene_expression_analysis`	Gene Expression Analysis	Bioinformatics	Normalization, transform, DE, PCA, clustering, QC
`glv_calibration`	gLV Calibration	Simulation	Fit a generalized Lotka-Volterra matrix from abundance
`glv_ecosystem`	gLV Ecosystem Simulation	Bioinformatics Analysis	Community dynamics via gLV ODE + perturbation + Monte Carlo
`glv_simulation`	gLV Simulation	Simulation	Perturbation on a pre-calibrated gLV matrix
`graph_miner`	GraphMiner	Machine Learning	Interpretable rules from Random Forests
`guild_ecology`	Guild Ecology	Ecological Modeling	Binary- vs CLR-tier co-occurrence per guild
`heatmap`	Heatmap	Visualization	Heatmaps with optional clustering
`join_datasets`	Join Datasets	Data Transform	INNER/LEFT/RIGHT/OUTER/CROSS/SEMI/ANTI joins
`line_chart`	Line Chart	Visualization	Line charts for time series / continuous data
`merge_node`	Merge	Control Flow	Merge data from multiple branches
`missing_value_handler`	Missing Value Handler	Data Transform	Impute missing values (incl. interpolation)
`network_inference`	Network Inference	Ecological Modeling	Signed taxon–taxon network; communities, centrality, edges
`network_viz`	Network Visualization	Ecological Modeling	Render network / DKG figures
`normalization`	Data Normalization	Data Transform	Scale numeric columns for ML prep
`otu_clustering`	OTU/ASV Clustering	Bioinformatics Analysis	Cluster amplicon sequences into OTUs / dereplicate ASVs
`pathway_enrichment`	Pathway Enrichment	Bioinformatics	Pathway enrichment (GO, KEGG, Reactome)
`pca_analysis`	PCA Analysis	Statistics	PCA with scree plot, biplot, diagnostics
`perturbation_predictor`	Community Perturbation Predictor	Bioinformatics Analysis	Predict cascading abundance changes on add/remove
`phylogenetic_tree`	Phylogenetic Tree Builder	Bioinformatics Analysis	Build a phylogenetic tree (Newick) from sequences
`power_analysis`	Microbiome Power Analysis	Bioinformatics Analysis	PERMANOVA-based sample-size / power via Monte Carlo
`render_figures`	Render Publication Figures	Visualization	Journal-styled publication figures
`representation_transform`	Representation Transform	Ecological Modeling	Transform abundance to binary / log1p / CLR
`scatter_plot`	Scatter Plot	Visualization	Scatter plots for correlation
`sequence_alignment`	Sequence Alignment	Bioinformatics	Needleman-Wunsch / Smith-Waterman / progressive MSA
`sequencing_depth`	Sequencing Depth Estimator	Bioinformatics Analysis	Recommend minimum sequencing depth
`statistical_summary`	Statistical Summary	Statistics	Summary statistics incl. outliers, confidence intervals
`switch_node`	Switch	Control Flow	Multi-way branch on a column value
`tabular_data_reader`	Tabular Data Reader	File Input	Read CSV/TSV/Excel (auto-detected)
`tabular_data_writer`	Tabular Data Writer	File Output	Write CSV/TSV/Excel (auto-detected)
`taxon_intervention_profile`	Taxon Intervention Profile	Ecological Modeling	Per-taxon intervention summary from a network envelope
`taxonomic_classification`	Taxonomic Classification	Bioinformatics Analysis	K-mer classification vs SILVA/GreenGenes/UNITE
`taxonomic_comparative_summary`	Taxonomic Comparative Summary	Bioinformatics Analysis	Comparative summary across samples (prevalence, core taxa)
`volcano_plot`	Volcano Plot	Visualization	Volcano plot: fold change vs significance