lossless-context-mcp

NORTHTEKDevs/lossless-context-mcp

★ 0 stars TypeScript 🤖 AI/LLM Updated today

MCP server that cuts agent file-read tokens ~72% without sacrificing quality — provably lossless (epoch-bounded dedup/diff).

View on GitHub → Try with Claude — $10 free →

Quick Install

Copy the config for your editor. Some servers may need additional setup — check the README.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "lossless-context-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "NORTHTEKDevs/lossless-context-mcp"
      ]
    }
  }
}

README Excerpt

An MCP server that **cuts the tokens agents spend re-reading files — without sacrificing quality.** On a realistic edit-loop coding workload it cuts file-read tokens **~72%**, measured with a real tokenizer, while staying **provably lossless**: across every read, what the model could reconstruct equalled the true file byte-for-byte (0 violations). On a read-once session it saves ~0% — the honest floor. Reproduce both with `npm run bench`.

Tools (1)

symbol

Topics

agentclaudeclaude-codecontextcost-optimizationllmmcptokens

lossless-context-mcp

Quick Install

README Excerpt

Tools (1)

Topics

Related AI/LLM Servers