Cache Proxy
Cut duplicate LLM spend in agent loops by routing calls through exact and semantic response caching paid per request via x402 USDC on Base.
Overview
io.github.Ainode-tech/cache-proxy is a Build-phase MCP server that fronts LLM APIs with exact and semantic response caching, x402 USDC billing on Base, and a free health check.
What is this MCP server?
- Streamable HTTP remote at cache.api.ainode.tech/mcp (MCP schema 2025-12-11)
- Exact plus semantic LLM response caching behind one proxy
- x402 micropayments in USDC on Base for paid cache hits
- Free health check endpoint for connectivity before you depend on it
- Version 0.1.1 published as io.github.Ainode-tech/cache-proxy
- Server version 0.1.1
- Two cache modes: exact and semantic
- Billing: x402 USDC on Base
What problem does it solve?
Agent workflows re-hit the same LLM prompts and burn budget because nothing deduplicates or semantically matches prior completions in your MCP stack.
Who is it for?
Solo builders running Claude Code or Cursor with heavy, repetitive LLM tool calls who already use or can enable x402 USDC on Base for pay-per-use APIs.
Skip if: Teams that need a fully self-hosted, flat-rate cache with no on-chain payment flow or who only make unique one-off prompts with no reuse pattern.
What do I get? / Deliverables
After you add the remote MCP URL, eligible agent requests can resolve from cache tiers so repeat work costs less and responds faster without custom cache code.
- Configured remote MCP entry pointing at cache.api.ainode.tech/mcp
- Verified free health check before production agent sessions
- Lower marginal LLM spend on repeated or near-duplicate prompts
Recommended MCP Servers
Journey fit
Solo builders wire MCP tools while assembling agents and APIs; caching belongs on the agent-tooling shelf where model integrations are configured. This server plugs into Claude Code, Cursor, and similar clients as a remote MCP endpoint—not a standalone app feature—so agent-tooling is the canonical placement.
How it compares
Remote LLM cache MCP integration, not an in-repo agent skill or local model runner.
Common Questions / FAQ
Who is io.github.Ainode-tech/cache-proxy for?
It is for indie and solo developers who connect MCP-capable agents to paid LLM APIs and want exact and semantic caching without building their own proxy.
When should I use io.github.Ainode-tech/cache-proxy?
Use it when your agent repeatedly asks similar questions or regenerates the same context and you are ready to pay per cached or uncached call via x402 USDC on Base.
How do I add io.github.Ainode-tech/cache-proxy to my agent?
Register the streamable-http remote MCP URL https://cache.api.ainode.tech/mcp in your client’s MCP config per vendor docs, verify the free health endpoint, then fund x402 USDC on Base as required by Ainode.