v1.1.7

Token optimization pass to keep more of the burden on GLM: trimmed all MCP tool descriptions + param docs (they sit in Claude's context every turn -> now ~1.5k tokens total), slimmed the hook's per-spawn injection (~500+ -> ~230 tokens), and glm_agent's concise default now omits the per-action list and full diff (Claude reads a short summary+stats, not GLM's whole output). Install: npx glm-mcp-claude@latest --key YOUR_ZAI_KEY