Clean JSON Toolkit

MCP Tools

Local JSON / JSONL inspection and manipulation toolkit. Inspect deeply nested structures (path tree, types, sample values per path), query with a jq-style path language (no external jq), flatten nested objects/arrays into dot-notation maps and unflatten them back, and validate against a small pragmatic schema (required/type/min/max/regex/enum/item_type). Pure Python 3 standard library, no third-party dependencies, no remote calls.

Install

openclaw skills install clean-json-toolkit

clean-json-toolkit

v0.1.0

Fourth member of the clean-* family. csv handles structured tabular data, text handles unstructured strings, log handles timestamped logs, and json handles nested data — API responses, config files, JSONL event streams.

Pure Python 3 standard library. No jq, no jsonschema, no pip installs.

Scripts

scripts/inspect.py — profile a .json or .jsonl file: tree of every distinct path with types, counts, and N sample values per path.
scripts/query.py — jq-style path queries. .key.nested, .key[0], .[] iterate, .key.[].field map. Output modes: --json / --jsonl / --lines / --raw. JSONL inputs implicitly iterate at the top level.
scripts/flatten.py — flatten nested JSON into dot-notation keys. Reversible with --unflatten. Roundtrip-safe.
scripts/validate.py — validate against a small schema (required, type, min/max, min_length/max_length, enum, regex, item_type, allow_extra).
scripts/check_deps.sh — verify python3.

Quick start

# Inspect
python3 scripts/inspect.py response.json
python3 scripts/inspect.py events.jsonl --max-samples 5

# Query
python3 scripts/query.py data.json '.meta'
python3 scripts/query.py data.json '.users.[].email' --raw
python3 scripts/query.py events.jsonl '.amount' --lines

# Flatten + unflatten roundtrip
python3 scripts/flatten.py config.json flat.json
python3 scripts/flatten.py flat.json nested.json --unflatten

# Validate
python3 scripts/validate.py users.jsonl --schema schema.json

Exit codes

Code	Meaning
0	success / one or more results
1	zero results / validation failed / empty input
2	bad arguments / unsafe path / missing file / invalid JSON / bad schema

Safety

Pure Python 3 stdlib. No eval, no subprocess, no remote calls.
All paths validated against safe-path policy (same as the other clean-* toolkits).
Hand-rolled query path tokenizer.

Pairs well with

License

MIT