Unity MCP Efficient v0.2.0

Unity MCP Efficient v0.2.0 is the first public release. It keeps MCP for Unity behind a small, stable interface so the model loads only the operation detail it needs.

Measured result

Against MCP for Unity v10.1.0, the model-visible surface changed from 48 tools to 6. Serialized tool schemas dropped from 93,119 to 3,657 characters: a 96.07% reduction, or about 23,280 to 915 tokens using the documented four-characters-per-token estimate. All 377 indexed Unity operations remain available through on-demand discovery.

These figures measure schema context, not total API billing. See the benchmark document for the exact baseline and stress-test limits.

Included

• Capability search with action-specific schemas
• Compact Unity-aware responses with recoverable raw-result handles
• Bounded batch workflows and conservative parallel reads
• Request receipts for safe retry handling
• Domain reload, stale status, nested failure, and asynchronous test handling
• Codex skill and task-specific reference guides

Verification

• 35 facade tests
• GitHub CI on Python 3.10, 3.11, 3.12, and 3.13
• Wheel and source distribution built from the release tag
• Live read-only smoke test against MCP for Unity v10.1.0

Install

Use the version-pinned commands in the README. Keep MCP for Unity running as the backend, and register only this facade with the AI client.

Artifact checksums

• unity_mcp_efficient-0.2.0-py3-none-any.whl: 01e1bfd59f9f9e00b0fd193e4f2a6a2b3be191f7532211f4905b7465a91839f0
• unity_mcp_efficient-0.2.0.tar.gz: 61e39fd33ae662582cd4b0e77795cc1c3b95bbc1b41595f04265c69ef74e7e1d

Install guide · Benchmarks · Changelog