# fp-conv

The fp-conv MCP server for Claude Code encodes, decodes, and converts floating-point and integer values. An agent can call it to inspect bit patterns or measure precision loss during conversion.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 8 repo stars
- Author: sw23
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/sw23/fp-conv/tree/HEAD/packages/fp-conv-mcp
- Updated: 2026-11-11
- Security check: automated, 2026-11-17 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/mcp/fp-conv

## What it does

fp-conv gives an agent tools to encode numbers, decode bit patterns, and move values between formats. Presets include FP32, FP16, BF16, TF32, FP8, FP6, FP4, and integer types. The server can also return format details or list the available presets.

Conversions support rounding and overflow options, and report precision loss. Custom format parameters are accepted too. Its HTTP mode runs locally without authentication; the standard connection command uses `npx -y fp-conv-mcp`.

## Requirements
- Needs: Node.js (commands run via npx)

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add fp-conv-mcp -- npx -y fp-conv-mcp
```

### Codex

**Platform CLI**
```bash
codex mcp add fp-conv-mcp -- npx -y fp-conv-mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "fp-conv-mcp": {
            "command": "npx",
            "args": [
                "-y",
                "fp-conv-mcp"
            ]
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "fp-conv-mcp": {
            "type": "stdio",
            "command": "npx",
            "args": [
                "-y",
                "fp-conv-mcp"
            ]
        }
    }
}
```

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add fp-conv-mcp npx -y fp-conv-mcp
```

## When to use it
- encode a value as BF16 and inspect its binary form
- decode an FP16 bit pattern into a decimal value
- convert a number from FP32 to FP8 and check precision loss
- look up parameters for available formats

## Example prompts
- Encode 3.14159 as BF16 and show its binary representation.
- Decode the bit pattern `0x3E00` as FP16.
- Convert 0.1 from FP32 to FP8 E4M3 and report the precision loss.

## Similar
- [Context7](https://affpapa.org/en/nevibecode/mcp/context7.md) — MCP server · 5563 installs
- [Search Google Flights](https://affpapa.org/en/nevibecode/mcp/search-google-flights.md) — MCP server · 2595 installs
- [Math-MCP](https://affpapa.org/en/nevibecode/mcp/math-mcp.md) — MCP server · 5958 installs
- [pipeworx gateway](https://affpapa.org/en/nevibecode/mcp/pipeworx-gateway.md) — MCP server · 419019 installs
- [SparkForge](https://affpapa.org/en/nevibecode/mcp/sparkforge.md) — MCP server · 219896 installs
- [OpenWebSearch](https://affpapa.org/en/nevibecode/mcp/openwebsearch.md) — MCP server · 494 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
