Skip to main content
MCP server CX Codex

Sophon — Claude Code MCP server: what it does, how to install

Sophon compresses MCP agent context, including chat history, prompts, command output, and file deltas. Use it when tool traffic keeps growing.

Install into Claude Code · Platform CLI All methods and platforms ↓
claude mcp add sophon -- npx -y mcp-sophon
Open on GitHub

What Sophon does

Sophon exposes MCP tools for reducing prompts, conversation history, and shell output. It can pass file edits as deltas, turn repeated fragments into token references, keep an incremental JSONL memory, and rank relevant parts of a repository for a code question.

The server ships as a single Rust binary and performs query-time compression without an ML model, GPU, or API key. Its output filters cover commands such as git, cargo, docker, kubectl, plus JSON data. You can run it through MCP with npx -y mcp-sophon.

There are practical limits. On raw source files, compress_prompt follows a budget cap instead of routing sections by the query. Some shell commands can add tokens, so the result depends on the data and the repository state.

What it helps with

  • compress a growing conversation history
  • trim git or cargo command output
  • send file changes instead of the full file
  • rank relevant code in a repository
  • replace repeated boilerplate with a token reference

Example prompts for the agent

  • Use compress_history to reduce this conversation while keeping its facts and recent context.
  • Run compress_output on the output from cargo test and retain the failures.
  • Use navigate_codebase to locate the code responsible for authentication.

How to install Sophon

Platform CLI

claude mcp add sophon -- npx -y mcp-sophon

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about Sophon

How do I install Sophon in Claude Code?

Run claude mcp add sophon -- npx -y mcp-sophon.

What does Sophon do?

It compresses dynamic MCP context such as prompts, conversation history, shell output, file changes, and code fragments. It also keeps JSONL memory and ranks repository content for code questions.

Does Sophon need an API key or GPU?

No. Query-time operation uses neither API keys nor a GPU or ML model; the server is built around a Rust binary.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor