kskurtveit/echocache

Overview

MCP server that caches expensive LLM results. Uses HTTP-style freshness to expire cached entries and semantic recall to retrieve related answers. Written in TypeScript.

Best for

Best for
Developers building MCP servers that need to cut LLM costs.

Use cases

• Cache repeated LLM calls with TTL-based freshness
• Retrieve semantically similar past answers for new queries
• Reduce cost and latency in MCP-based LLM workflows

How to use

Install

claude mcp add echocache -- npx -y echocache

Tools exposed

• ECHOCACHE_DB_PATH
• ECHOCACHE_MAX_ENTRIES
• ECHOCACHE_MAX_BYTES
• ECHOCACHE_DEFAULT_TTL_SECONDS
• ECHOCACHE_SIMILARITY_THRESHOLD
• ECHOCACHE_LINK_CANDIDATE_POOL
• ECHOCACHE_ENCRYPTION_KEY
• cache_get
• cache_set
• cache_query
• cache_related
• cache_invalidate
• cache_stats

Tested with

Claude Desktop, Claude Code, Cursor, VS Code

Notes

MCP server that caches expensive LLM results. Uses HTTP-style freshness to expire cached entries and semantic recall to retrieve related answers. Written in TypeScript.

0 stars on GitHub. Last updated 2026-09-10. Licensed MIT.

Use cases

• Cache repeated LLM calls with TTL-based freshness
• Retrieve semantically similar past answers for new queries
• Reduce cost and latency in MCP-based LLM workflows

Pros

• Simple HTTP-style caching model
• Semantic recall improves hit rate for paraphrased queries
• Lightweight TypeScript implementation

Cons

• Requires running and maintaining a separate MCP server
• Semantic recall may need embedding infrastructure
• No stars or community adoption yet

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.