Bio-Specific Usage Tips

Pipeline Generation Tips

Be Specific with Intent

The more specific your natural language description, the better the pipeline:

# Good — specific workflow with clear steps
bc7 create "RNA-seq from paired-end FASTQ to differential expression using STAR and DESeq2" \
  --target nextflow
 
# Less specific — may produce a more generic pipeline
bc7 create "RNA-seq pipeline" --target nextflow

Specify Container Preferences

Force container resolution for reproducibility:

bc7 create "variant calling pipeline" \
  --target nextflow \
  --containers docker

Use EDAM Terms for Precision

When searching, EDAM terms give more precise results than free text:

# EDAM operation search (precise)
bc7 search --operation "operation:0292"  # Sequence alignment
 
# Free text (broader)
bc7 search "alignment"

Registry Query Tips

Combine Filters

# Search by operation AND topic
bc7 search --operation alignment --topic genomics
 
# Search with format constraints
bc7 search --operation alignment --input-format fastq --output-format bam

Explore EDAM Hierarchy

EDAM is hierarchical — broader terms include narrower ones:

# See child terms
bc7 edam children "operation:0292"
 
# See parent terms
bc7 edam parents "operation:0292"

MCP Integration Tips

Prompt Patterns for Claude Code

When using BioContext7 through Claude Code's MCP integration, these prompt patterns work well:

• "Search for tools that can perform sequence alignment on FASTQ files"
• "Suggest a pipeline for RNA-seq differential expression analysis"
• "What operations does the STAR aligner support?"
• "Look up the EDAM term for variant calling"

Single-Call with find-skills (Recommended)

For most use cases, find-skills is the fastest path — it returns ranked tools with install commands, quality scores, and health signals in a single call:

find-skills(task_description="variant calling for whole genome sequencing", compact=true)

Set compact: true (the default) for LLM-optimized responses that minimize token usage. Use the platform parameter to filter install commands for your target system:

find-skills(task_description="RNA-seq alignment", platform="darwin-arm64")

Chaining MCP Calls (Two-Step Pattern)

For detailed documentation on a specific tool, chain resolve-library-id and get-library-docs:

1. resolve-library-id("STAR") — find matching tool IDs
2. get-library-docs(id="star", topic="2-pass alignment", maxTokens=3000) — fetch targeted docs

Use get-library-docs with mode="document" for full source documents instead of ranked snippets.

Relevance Feedback

After getting documentation snippets, use report-snippet-quality to improve future results:

report-snippet-quality(chunk_id=42, helpful=false)

Unhelpful snippets are penalized in subsequent retrievals within the same session.

Performance Tips

Local Index

BioContext7 maintains a local index for fast searches. Keep it updated:

# Full sync (may take a few minutes)
bc7 sync all
 
# Quick sync (only changed tools since last sync)
bc7 sync --incremental

Caching

API responses are cached locally by default. Clear the cache if you need fresh data:

bc7 cache clear

Common Pitfalls

Missing EDAM Annotations

Some tools in bio.tools lack EDAM annotations. BioContext7 uses text-based matching as a fallback, but results may be less precise. Check tool details with bc7 info <tool> to verify annotations.

Container Availability

Not all tools have Docker/Singularity containers available. BioContext7 checks BioContainers automatically, but for tools without containers you may need to install them manually.

Pipeline Target Limitations

Each compilation target has different capabilities: