Back to directory
Frontendwesleyscholl/squish
squish-ai
Local LLM inference server for Apple Silicon. Block-level paged KV cache for long-context workloads. 5.4× faster end-to-end on 4K-token prompts vs Ollama, less RAM, INT3 support for Qwen3. OpenAI-compatible API.
Suggested install command
npx skills add wesleyscholl/squish/squish-aiAlways inspect the linked repository and skill instructions before running commands. Skills are instructions; permissions and execution still matter.
Compatibility
Agent support matrix
3 supported
| Agent | Status |
|---|---|
| Claude Code | Supported |
| OpenCode | Not listed |
| Cursor | Supported |
| MCP | Not listed |
| GitHub Copilot | Not listed |
| Windsurf |