token-optimizer-mcp

An MCP server that slashes AI agent token costs by compressing context in a cache-aware way — it preserves prompt-prefix boundaries so reused tokens re-read at 0.1x instead of being billed at 1.25x.

Connects to: Claude Code, Cursor, VS Code, 16 MCP-compatible clients · · MIT 536★

Use it with an AI agent

Loadbay is an MCP server, so an agent can search the catalog and find this harness:

claude mcp add --transport http loadbay https://loadbay.xyz/api/mcp

• Source: https://github.com/ooples/token-optimizer-mcp
• This harness as JSON: /api/harnesses/token-optimizer-mcp
• Agent setup: /setup.md
• Browse all 850+ harnesses on Loadbay