Use custom channels

By default, conda-exec searches conda-forge for packages. You can specify alternative or additional channels from the CLI or inside script metadata.

Specify a channel from the CLI

Use -c or --channel to search a different channel:

conda exec -c bioconda samtools view input.bam

Use multiple channels

Repeat -c to add multiple channels. They are searched in the order given:

conda exec -c bioconda -c defaults samtools view input.bam

Important

When you pass -c, the default conda-forge channel is not added automatically. If your packages depend on conda-forge, you must include it explicitly.

Include conda-forge alongside other channels when needed:

conda exec -c conda-forge -c bioconda samtools view input.bam

Bioinformatics example

Many bioinformatics tools live in the bioconda channel and depend on packages from conda-forge. Specify both:

conda exec -c conda-forge -c bioconda minimap2 -a ref.fa reads.fq > out.sam

Declare channels in script metadata

For Python scripts with PEP 723 inline metadata, declare channels in the [tool.conda] table:

# /// script
# [tool.conda]
# channels = ["conda-forge", "bioconda"]
# dependencies = ["samtools>=1.19", "pysam"]
# ///

import pysam
print(pysam.__version__)

conda exec script.py

Combine script channels with CLI channels

Tip

Channels are searched in order. Channels declared in script metadata come first, followed by any CLI channels. Put your preferred channel first to give it higher priority.

CLI channels (-c) are appended after channels declared in script metadata. Given a script that declares channels = ["conda-forge"], running:

conda exec -c bioconda script.py

resolves packages using ["conda-forge", "bioconda"] in that order.

If neither the script metadata nor the CLI provides channels, conda-forge is used as the default.

How channels affect caching

Different channel combinations produce different cache keys. Running the same tool with different -c flags creates separate cached environments:

conda exec ruff check .                 # cached as ruff--<hash1>
conda exec -c defaults ruff check .     # cached as ruff--<hash2>