nxp-monkey 2026.6.9

Cross-platform CLI and Python library for NXP MCUXpresso Config Tools chip data

                    ██████████
                  ██▓▓▓▓▓▓▓▓▓▓██
                ██▓▓░░░░░░▓▓▓▓░░██
              ████░░      ░░░░    ██
            ██░░▓▓░░  ████░░░░██  ██
            ██░░▓▓░░  ████░░░░██  ██
              ██▓▓░░  ░░░░░░░░░░██
              ██▓▓▓▓▓▓▓▓▓▓░░▓▓░░▓▓██
    ████        ██▓▓▓▓▓▓▓▓░░░░░░▓▓██
  ██▓▓▓▓██    ██▓▓████████▓▓▓▓▓▓██
██▓▓████      ██▓▓▓▓▓▓▓▓▓▓████████
██▓▓██      ██▓▓▓▓██▓▓██▓▓▓▓▓▓▓▓██
██▓▓██    ████▓▓▓▓██▓▓▓▓██████▓▓██
██▓▓██████▓▓██▓▓▓▓██▓▓▓▓▓▓██▓▓▓▓██
  ██▓▓▓▓▓▓████░░░░██░░░░░░██░░░░██
    ██████    ████████████  ████

nxp_monkey

Cross-platform CLI and Python library for fetching, indexing, and searching NXP MCUXpresso Config Tools chip data, with both XML (raw NXP binders) and JSON (parsed spine + lossless XML->JSON mirror) output by default.

nxp_monkey talks to NXP's public KEX storage API (the same one the MCUXpresso Config Tools Data Manager uses) and exposes the per-processor data tree — signal configurations, register variants, packages, clocks, resource tables — as a clean local cache that humans, scripts, and LLM agents can all consume.

Install

uv tool install nxp-monkey
uv tool update-shell
nxp-monkey --version
nxp-monkey version
# Short alias is also installed:
nxpm --version
nxpm version

CLI quick start (MCXA156)

# What tool versions does NXP publish?
nxp-monkey versions

# What processor families are available?
nxp-monkey families

# Search by part / family root. Uses the index when present and falls
# back to the portfolio map when it is not.
nxp-monkey search MCXA
nxp-monkey search mcxa --fuzzy

# Build a local index of all available parts and variants
nxp-monkey index build
nxp-monkey index show MCXA156

# Fetch one part. Downloads every SDK variant NXP publishes,
# mirrors the XML, and writes the JSON views alongside.
nxp-monkey fetch MCXA156

# Prefixes and concrete orderable aliases are accepted too.
nxp-monkey fetch MIMX93 --variant ksdk2_0

# Print the parsed chip-data spine (header, cores, package variants)
nxp-monkey details MCXA156

# Print the per-part roadmap (where things live + inferred schema)
nxp-monkey roadmap MCXA156

# Manage the cache
nxp-monkey cache path
nxp-monkey cache size
nxp-monkey cache clear

What you get from fetch MCXA156

Output is split by media type so an agent can pick either side:

MCXA156/
  xml/                              # raw NXP binders, one subdir per SDK variant
    ksdk2_0/...                     #  - canonical silicon data (registers, pins, clocks)
    zephyr3_2/...                   #  - Zephyr DT codegen scripts
    i_mx_2_0/...                    #  - i.MX Linux binder (when published)
  json/
    MCXA156.json                    # PartDetails spine: header, cores, package SKUs, db_links
    MCXA156.roadmap.json            # agent guide + inferred schema + folder layout
    ksdk2_0/...                     # full XML->JSON mirror (one .json per .xml)
      MCXA156VFT/registers/ADC1.json
      MCXA156VFT/signal_configuration.json
      packages/QFN48.json
      ...

For MCXA156 this is ~40 MB of XML plus a 1:1 ~71 MB JSON mirror across 127 files (per variant). Progress is shown live on stderr; stdout stays a clean list of paths so it pipes well.

JSON vs XML

You want...	Read
Top-level chip facts (family, cores, packages, SKU list)	json/<PART>.json
A guide to where things live + observed XML namespaces	json/<PART>.roadmap.json
One peripheral's registers / bit fields	json/<variant>/<SKU>/registers/<PERIPH>.json
Pin mux for a specific SKU	json/<variant>/<SKU>/signal_configuration.json
The bit-identical NXP binder (codegen scripts, exact whitespace)	xml/<variant>/...

The XML->JSON conversion is lossless for data: namespace-stripped tags, @attr keys for attributes, #text for mixed content, repeated children as lists, root @_xmlns preserves the source namespace declarations. Skip the mirror with --no-json-mirror, or narrow it with --json-mirror-only '**/registers/**' (repeatable glob).

The global --json flag switches stdout to print the JSON output path (for tool / agent integration); JSON files are always written by fetch.

Library quick start

import nxp_monkey

# Discovery
versions = nxp_monkey.list_versions()
families = nxp_monkey.list_families()

# Search
hits = nxp_monkey.search("MCXA", fuzzy=False)
imx_hits = nxp_monkey.search("MIMX9352CVVXMAB")

# Index
nxp_monkey.build_index()
info = nxp_monkey.get_part("MCXA156")     # -> PartInfo dataclass

# Fetch + parse
path = nxp_monkey.fetch("MIMX9352CVVXMAB") # -> pathlib.Path to unpacked tree
details = nxp_monkey.details("MCXA156")   # -> PartDetails (header, cores, variants)
roadmap = nxp_monkey.build_roadmap(path)  # -> dict (guide, key_files, xml_namespaces, ...)

# Cache control
nxp_monkey.cache_path()
nxp_monkey.cache_size()
nxp_monkey.cache_clear()

Documentation

• docs/setup.html - setup, release, and artifact policy.
• docs/architecture.html - package layers and ownership boundaries.
• docs/adrs/ — architecture decision records.
• docs/design/cli/ — one HTML doc per CLI command.
• docs/design/api/ — one HTML doc per public library interface.
• docs/contracts/ — JSON manifests + schemas (command + interface).
• docs/research/xml_survey.md — narrative survey of the KEX XML schema universe.

Testing

Rack is the primary local signoff gate:

uv sync --all-extras
uv run rack run --all
uv run python -m build
uv run twine check dist/*
uv run python tests/support_scripts/install_test.py

License

MIT. See LICENSE.

nxp_monkey fetches public NXP data on behalf of the user. NXP MCUXpresso Config Tools content is NXP material and is governed by NXP terms.