board.GetLayerStack() was removed in KiCad 9. Call SetLayerName and
SetLayerType directly on the board instead, and grow the copper layer
count via SetCopperLayerCount when adding inner layers. Without this,
add_layer raises AttributeError on any KiCad 9 installation.
Promotes BOM / sourcing fields (MPN, Manufacturer, DigiKey_PN, LCSC,
JLCPCB_PN, Voltage, Tolerance, Dielectric, ...) to first-class citizens
on placed schematic symbols.
New MCP tools:
- set_schematic_component_property: add or update one custom property
on a component (convenience wrapper around edit_schematic_component).
- remove_schematic_component_property: delete one custom property.
The four built-in fields (Reference, Value, Footprint, Datasheet) are
protected and rejected.
edit_schematic_component enhancements:
- New `properties` parameter: map of property name to either a string
value or a full spec object { value, x?, y?, angle?, hide?, fontSize? }.
Adds the property when missing, otherwise updates the existing field
(and optionally its label position / visibility). Lets a single tool
call attach an entire BOM payload to a component.
- New `removeProperties` parameter: list of custom property names to
delete in the same call.
- Property values are now backslash-escaped so descriptions containing
a double-quote or a backslash no longer corrupt the .kicad_sch file.
- New properties default to (hide yes) so they appear in BOM exports
without cluttering the schematic canvas.
get_schematic_component description clarified to highlight that it
already returns every field on the symbol, including custom ones.
New MCP prompt component_sourcing_properties guides agents through the
conventional property names recognised by downstream BOM tooling and
the recommended call sequence.
Implementation (python/kicad_interface.py):
- _PROTECTED_PROPERTY_FIELDS frozenset
- _escape_sexpr_string / _find_matching_paren static helpers
- _set_property_in_block / _set_hide_on_property /
_remove_property_from_block surgical text-level edits that preserve
formatting and the property's UUID
- _handle_edit_schematic_component rewritten to orchestrate
add/update/remove and return a per-property summary
- New handlers _handle_set_schematic_component_property and
_handle_remove_schematic_component_property registered in the
command dispatch table
Tests (tests/test_schematic_component_properties.py):
32 tests covering escape helper, paren matcher, add/update/remove
(single + batched), full spec dicts, default position, default
(hide yes), special-character escaping, UUID preservation, protected
built-in field rejection, no-op removal, both new convenience tools,
and input validation. All 590 tests in the project still pass.
Docs: README, SCHEMATIC_TOOLS_REFERENCE, TOOL_INVENTORY, CHANGELOG.
Adds robust multi-sheet (hierarchical) net connectivity for KiCad
schematics and switches the wire/label parsing to a direct sexpdata
pipeline that bypasses kicad-skip's collection iteration, which was
silently dropping wires, labels, and symbol instances on real-world
schematics.
python/commands/wire_connectivity.py
- New sexpdata helpers: _load_sexp, _parse_wires_sexp,
_parse_labels_sexp, _parse_symbol_instances_sexp,
_parse_hierarchical_labels_sexp, _discover_sub_sheets.
- _build_adjacency now detects T-junctions (endpoint landing on
another wire's interior segment) so adjacency captures connections
KiCad doesn't represent as separate wire segments.
- _find_connected_wires gains an interior-segment fallback so labels
placed mid-wire still seed BFS correctly.
- _parse_virtual_connections gathers label / global_label /
hierarchical_label and power-symbol pin positions, with a
kicad-skip fallback for unit tests that mock the schematic.
- _find_pins_on_net rebuilds pin positions from sexpdata symbol
instances (with mirror_x/mirror_y/rotation handling) and uses a
plus/minus 1 IU tolerance for floating-point edge cases.
- get_connections_for_net walks the top sheet plus every recursively
discovered sub-sheet, deduping pins across sheets.
python/commands/pin_locator.py
- lib_id matching now falls back to a bare-name + unit-suffix match
so instances like "stat-tis-custom:BAT_18650" resolve to
lib_symbols entries like "BAT_18650_3".
- Pin position math now y-negates lib_symbols coords, applies
mirror_x/mirror_y in local coords before rotation, and propagates
the same transform into get_pin_orientation so downstream callers
get a correct outward angle for mirrored symbols.
python/commands/connection_schematic.py
- generate_netlist now collects nets from both label and
global_label and routes them through get_connections_for_net so
netlists reflect cross-sheet connectivity instead of single-sheet
label-only nets.
python/kicad_interface.py
- list_schematic_nets aggregates net names across the top sheet and
all sub-sheets via the sexp helpers, then resolves connections
using get_connections_for_net.
- get_net_connections delegates to get_connections_for_net for
consistent multi-sheet results.
Three bugs fixed in the schematic component and pin locator pipeline:
1. component_schematic: remove redundant symbol.append() after clone()
kicad-skip's clone() already inserts the raw element into the schematic
tree. The subsequent NamedCollection.append() detects the reference as
already registered (from the elementRename triggered by setting
property.Reference.value) and renames it "R1_" with a trailing
underscore, causing all subsequent pin lookups to fail.
2. pin_locator: negate lib y coordinate before rotation
lib_symbols in .kicad_sch use library y-up convention; schematic
coordinates use y-down. get_pin_location now negates pin_rel_y before
applying rotation, matching KiCad's own transform order (same approach
as _transform_local_point in schematic_analysis.py).
3. pin_locator: add .rstrip("_") guard in all symbol reference lookups
Defensive guard against any residual cases where kicad-skip writes a
trailing underscore to the Reference property value.
Also fixes the self-test script to use template_with_symbols.kicad_sch
(which contains placed _TEMPLATE_* symbols) rather than the expanded
template (which only contains lib_symbols definitions and has no cloneable
instances).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds two new MCP tools for working with free-form text annotations
(SCH_TEXT elements) in KiCad schematics:
- add_schematic_text: place a text note with optional angle, font size,
bold/italic, and justification
- list_schematic_texts: list all text annotations with optional
case-insensitive substring filter
Includes WireManager.add_text / list_texts using _text_insert + sexpdata,
handler dispatch in KiCADInterface, TypeScript tool definitions, registry
entry, reference doc updates, and 30 unit tests.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
15 tests covering:
- Empty schematic → empty maps
- Global / local / hierarchical labels at pin endpoints
- Net propagation through one and two wire hops
- Unconnected pin stays absent from map
- Power (#PWR) and flag (#FLG) symbols excluded from component map
- Components in sub-sheets collected alongside top-level components
- 90° rotated symbol produces correct absolute pin positions
Uses MagicMock skip.Schematic objects patched in both the walker
import and commands.pin_locator module namespace; lib_symbols are
provided in real on-disk .kicad_sch files so sexpdata parsing runs
without mocking.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Hierarchical KiCad schematics store all components in sub-sheets;
the top-level .kicad_sch only contains sheet references. The previous
sync_schematic_to_board implementation called generate_netlist on the
top-level file only, which has no components, so it always returned
0 pads assigned.
Replace with _build_hierarchical_pad_net_map which:
- rglobs all .kicad_sch files in the project directory
- For each sheet, collects label positions from label, global_label,
and hierarchical_label via skip.Schematic
- Adds power symbol (#PWR/#FLG) positions using their Value as net name
- Builds a wire adjacency graph and BFS-propagates net names through
wire segments to reach pins not directly under a label
- Calls PinLocator.get_all_symbol_pins to get absolute pin positions,
then matches to the propagated net map within 0.5 mm tolerance
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds python/annotations/loader.py (AnnotationLoader) which resolves MCP
tool names to KiCad IPC proto message annotations using three layers:
1. Explicit TOOL_TO_PROTO mapping for non-obvious name pairs
2. Automatic snake_case → PascalCase conversion
3. Suffix-stripped variants (get_nets_list → GetNets)
Integrates into kicad_interface.py tools/list handler:
- Tools with existing schemas: enrich_schema() adds blocking/interactive
ToolAnnotations hints and fills description gaps
- Tools without schemas (fallback): gets a real description from the
IPC annotation instead of the generic "KiCAD command: <name>" string
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Generated by scripts/generate_tool_annotations.py --fetch-from-gitlab --use-cli.
Covers all command messages across board, schematic, and common proto files.
Stored in python/annotations/ (data/ is gitignored for large JLCPCB databases).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds a second annotation backend that shells out to the 'claude' CLI
(Claude Code) instead of calling the Anthropic SDK. This allows users
with a Claude.ai monthly subscription to run the annotation script
without needing an API key.
The SDK backend (default) still uses prompt caching on the proto context
block for cheaper repeated/resumed runs. The CLI backend skips caching
but works out of the box with a Claude.ai account.
Usage:
# Monthly plan (no API key)
python scripts/generate_tool_annotations.py --fetch-from-gitlab --use-cli
# API key (with prompt caching)
python scripts/generate_tool_annotations.py --fetch-from-gitlab
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Reads KiCad's IPC API proto definitions and uses the Claude API to generate
rich, user-facing descriptions for MCP tool metadata. The output JSON file
can be loaded by an MCP server at startup to annotate auto-generated tools.
Features:
- Fetches proto files from KiCad GitLab or loads from a local checkout
- Pure-Python proto parser (no protoc dependency)
- Prompt caching on the static proto context block — re-runs are cheap
- --resume to skip already-annotated messages after an interrupted run
- --dry-run to preview what would be annotated without calling the API
- Reports cache hit/write/read token counts after each API call
Usage:
python scripts/generate_tool_annotations.py --fetch-from-gitlab
python scripts/generate_tool_annotations.py --proto-dir /path/to/kicad/api/proto
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The Python executable and PYTHONPATH were hardcoded to Windows paths,
causing the server to fail silently on macOS/Linux. Now reads
KICAD_PYTHON from the environment (set by setup-macos.sh) with
python3 as the fallback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
WireManager.delete_label only checked for Symbol("label") when scanning
the schematic s-expression list, so it silently skipped and failed to
delete global_label and hierarchical_label elements — returning False
with "No matching label found" even when a visible label existed at the
given coordinates.
Fix: collect all three label-like symbol types into a set and use `in`
instead of `==` for the type check.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Covers create_component_instance and _handle_add_schematic_component:
- default unit=1 behaviour
- explicit unit 2 and 4 written to (unit N) in schematic S-expression
- (instances ...) block also records the correct unit
- multiple units of the same reference placed independently
- handler param validation (missing schematicPath, missing component)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Allows placing a specific unit (A=1, B=2, C=3, …) of a multi-unit KiCad
symbol rather than always defaulting to unit 1. Required for quad
optocouplers, dual op-amps, and other multi-unit parts where each channel
must be placed independently.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Covers the three bugs fixed in the previous commit:
- TestAddComponentNoTrailingUnderscore (integration): verifies clone() + no extra append()
- TestPinLocatorYAxisNegation (unit): rotation=0 and 90° cases with y-negation
- TestPinLocatorReferenceRstrip (unit): lookup tolerates 'R1_' artifact from kicad-skip
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Three bugs fixed in the schematic component and pin locator pipeline:
1. component_schematic: remove redundant symbol.append() after clone()
kicad-skip's clone() already inserts the raw element into the schematic
tree. The subsequent NamedCollection.append() detects the reference as
already registered (from the elementRename triggered by setting
property.Reference.value) and renames it "R1_" with a trailing
underscore, causing all subsequent pin lookups to fail.
2. pin_locator: negate lib y coordinate before rotation
lib_symbols in .kicad_sch use library y-up convention; schematic
coordinates use y-down. get_pin_location now negates pin_rel_y before
applying rotation, matching KiCad's own transform order (same approach
as _transform_local_point in schematic_analysis.py).
3. pin_locator: add .rstrip("_") guard in all symbol reference lookups
Defensive guard against any residual cases where kicad-skip writes a
trailing underscore to the Reference property value.
Also fixes the self-test script to use template_with_symbols.kicad_sch
(which contains placed _TEMPLATE_* symbols) rather than the expanded
template (which only contains lib_symbols definitions and has no cloneable
instances).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Moves a net label (local, global, or hierarchical) to a new position in
place, avoiding the error-prone delete-then-re-add workflow. Supports an
optional currentPosition disambiguator and labelType filter.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The Python executable and PYTHONPATH were hardcoded to Windows paths,
causing the server to fail silently on macOS/Linux. Now reads
KICAD_PYTHON from the environment (set by setup-macos.sh) with
python3 as the fallback.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Bundles two changes for python/commands/schematic.py:
1. CRLF → LF line-ending normalization (matches the bulk renormalize
from the previous commit — was held back here because black would
also need to re-reformat it).
2. Black reflow of create_schematic()'s parameter list, which exceeded
the line length. Pre-existing drift, no logic change.
`git show --ignore-all-space HEAD` shows just the 2-line signature diff.
Mechanical application of the `.gitattributes` rules from the prior commit.
All 50 files differ only in line endings — verified by
`git diff --cached --ignore-all-space` being empty.
Before: main had 42 CRLF + 27 LF Python files plus mixed-ending in YAML,
templates, and shell scripts. After: every text file is LF (except the
Windows-native *.ps1, *.bat scripts which remain CRLF per gitattributes).
This eliminates the noisy-diff failure mode seen in PR #102, where a
small logic change produced a 918-line diff due to whole-file CRLF→LF
conversion.
Introduces a three-layer defence against CRLF/LF drift:
- `.gitattributes` normalizes all text files to LF on commit and checkout;
Windows-native scripts (*.bat, *.cmd, *.ps1) explicitly keep CRLF.
- `.editorconfig` instructs editors (VS Code, JetBrains, vim, etc.) to save
with LF by default so the problem is prevented at the source.
- `mixed-line-ending --fix=lf` hook in pre-commit catches anything that
slips through and auto-fixes it before commit.
No file contents are normalized in this commit — that follows in a
separate `git add --renormalize .` commit for easier review.
The placeholder static method on LibraryManager (library_schematic.py)
returned [] and had no dispatch entry or caller. The real tool of the
same name lives in library_symbol.py.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add optional netName (exact case-sensitive match) and labelType
(net/global/power enum) parameters. Both are optional and AND
together when combined. Omitting both preserves current behaviour.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Parser used a 5000-char heuristic slice per symbol; any symbol block
shorter than 5000 chars bled into the next one, and last-write-wins in
the properties dict ensured the neighbor's data clobbered the target.
Reported as Simulation_SPICE:OPAMP returning PJFET data.
Switch to parenthesis-depth tracking to find the true end of each
(symbol ...) block. Also surface Sim.Pins so agents can read opamp pin
numbering without inferring it from schematic placement.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
