README.md
A Model Context Protocol (MCP) server for TRPG dice rolling
Dice MCP Server
A Model Context Protocol (MCP) server for TRPG dice rolling across multiple game systems.
Quick Start
uvx (recommended)
{
"mcpServers": {
"dice": {
"command": "uvx",
"args": ["roleplaying-dice-mcp"]
}
}
}
Config: macOS
~/Library/Application Support/Claude/claude_desktop_config.json/ Windows%APPDATA%\Claude\claude_desktop_config.json
pip / Docker
pip install roleplaying-dice-mcp
roleplaying-dice-mcp
docker build -t dice-mcp-server . && docker run -i dice-mcp-server
Tools
| Tool | Purpose | Key Params |
|---|---|---|
roll_dice |
Roll dice using standard notation (NdM+X, kh/kl, dF) | notation, advantage, bonus_dice, target, degrees |
roll_pool |
Dice pool with success counting (WoD, Shadowrun) | pool, target, explode, count_ones |
reroll |
Re-roll last roll with same params | — |
get_history |
Retrieve recent roll log | limit |
clear_history |
Clear roll history | — |
See CONTEXT.md for system-specific parameter mappings and invariants.
Parameter Reference
`roll_dice`
| Parameter | Type | Default | Description |
|---|---|---|---|
notation |
string | required | Dice expression: 1d20+5, 4d6kh3, 4dF+2, 1d100 |
advantage |
bool | false |
Roll d20 twice, take higher (1d20 only) |
disadvantage |
bool | false |
Roll d20 twice, take lower (1d20 only) |
bonus_dice |
int | 0 |
CoC bonus dice (0–2, 1d100 only) — extra tens, keep lowest |
penalty_dice |
int | 0 |
CoC penalty dice (0–2, 1d100 only) — extra tens, keep highest |
target |
int | — | DC/AC/skill value for success check |
target_mode |
enum | "at_least" |
"at_least" (D&D: ≥) or "at_most" (CoC: ≤) |
critical |
bool | false |
Nat 20/1 detection. With degrees="pf2e", shifts degree ±1 step |
degrees |
enum | — | "coc" / "pf2e" / "pbta" — enables degree calculation |
`roll_pool`
| Parameter | Type | Default | Description |
|---|---|---|---|
pool |
int | required | Number of dice (1–50) |
sides |
int | 10 |
Sides per die |
target |
int | 8 |
Minimum value for a success (≥) |
explode |
bool | false |
Reroll on max value |
double_on |
int | — | Double success at this value or higher (must be ≥ target) |
count_ones |
bool | false |
Enable botch/glitch detection |
`get_history`
| Parameter | Type | Default | Description |
|---|---|---|---|
limit |
int | 10 |
Number of recent rolls to retrieve (1–100) |
System Coverage
| System | Coverage | Key Features Used |
|---|---|---|
| D&D 5e | 95% | notation, advantage, critical, target, 4d6kh3 |
| CoC 7e | 85% | d100, bonus/penalty dice, degrees=coc, at_most |
| PF2e | 85% | d20+mod, degrees=pf2e, critical (nat 20/1 shift) |
| PbtA | 90% | 2d6+mod, degrees=pbta |
| FATE | 95% | 4dF+mod |
| WoD | 90% | roll_pool, explode, count_ones (botch) |
| Shadowrun | 85% | roll_pool (d6), count_ones (glitch) |
| Savage Worlds | 50% | Exploding works, but no wild die or raise counting |
Project Structure
├── server.py # MCP server — tools & handlers
├── dice_parser.py # Notation parser (NdM, NdF, kh/kl)
├── history.py # Roll history manager
├── src/roleplaying_dice_mcp/ # PyPI package
├── tests/test_trpg_scenario.py
├── pyproject.toml
├── Dockerfile
└── README.md
License
MIT
Tools (5)
roll_diceRoll dice using standard notation (NdM+X, kh/kl, dF).roll_poolDice pool with success counting (WoD, Shadowrun).rerollRe-roll last roll with same params.get_historyRetrieve recent roll log.clear_historyClear roll history.Configuration
claude_desktop_config.json
{"mcpServers": {"dice": {"command": "uvx", "args": ["roleplaying-dice-mcp"]}}}Try it
→Roll 1d20 with advantage for my D&D attack check.
→Roll 5d10 for a Shadowrun dice pool with success counting.
→What was the result of my last three dice rolls?
→Roll 4dF+2 for my FATE character action.
→Perform a Call of Cthulhu skill check with 1 bonus die.