Clash of Clans MCP Server
Enables AI-assisted clan analysis for Clash of Clans by reading official API data and applying customizable rubrics for war grading, promotion screening, and carry-forward decisions.
README
Clash of Clans MCP Server
A Model Context Protocol (MCP) server that lets Claude (or any MCP client) read your clan's data from the official Clash of Clans API and apply your clan's own rules to it — for war grading, elder promotion screening, and CWL carry-forward decisions.
Built for clan leaders who want AI-assisted clan analysis with reproducible, configurable rules.
Keywords: Clash of Clans, MCP, Model Context Protocol, Claude, Claude Desktop, clan war, CWL, Clan War League, clan management, elder promotion, war analysis, Supercell API.
What you can do with it
Once connected to Claude Desktop (or another MCP client), you can ask things like:
"How did our last war go? Who didn't follow the rules?"
"Should I promote ankit176 to Elder? What's our policy bar?"
"Pull together a CWL carry-forward recommendation for next season."
"Who in the clan has the highest donation ratio this season?"
"Pull our capital raid loot for the last 5 weekends."
The MCP turns each question into the right API calls, applies your configurable rubric (in JSON, edit freely), and returns a leadership-ready answer.
Sample output
After you run it on a finished war, you get something like:
# War Report — vs clash serious (Defeat)
Total attacks used: 26 · Missed: 4 · Avg ⭐/attack: 2.27 · Avg destruction: 83%
## 1. Attack Participation (2/2 expected)
- Used both attacks (12): Justin (#1), Slayer7 (#2), TURBO (#3), ...
- Used only 1 of 2 (2): M@NJ!T (#4), Laeo_Lol XD (#11)
- 🚨 Did not attack (1): Punjabi Pete (#15)
## 2. Rule Compliance
- Clean (3): Justin (#1), aNkiT (#6), joe (#10)
- Violations (12):
- Slayer7 (#2, TH18): 2nd attack didn't 3⭐ and only hit 80%
- M@NJ!T (#4, TH18): 1st attack hit base #10 (offset +6); rule = mirror or one down
- ankit176 (#8, TH17): 2nd attack target offset -3 (should hit lower bases for smart 3⭐)
- ...
## 3. Performance Leaderboard
| Rank | Name | Pos | TH | ⭐ Earned | Avg Dest % | Score | Grade |
| 1 | Justin | #1 | TH18 | 6 | 100% | 23 | A |
| 2 | Corrupt | #5 | TH17 | 6 | 100% | 20 | A |
...
## 4. Smart-Attack Honor Roll (clean 3⭐ on lower base)
- Justin (#1) → hit base #2 for 3⭐ 100%
- Corrupt (#5) → hit base #9 for 3⭐ 100%
Tools the MCP exposes
API wrappers (read-only):
| Tool | Purpose |
|---|---|
clash_get_clan |
Clan-level details (name, level, war league, members count) |
clash_get_clan_members |
Roster with role, donations, war stars, trophies |
clash_get_warlog |
Past wars (summary only — see limitations) |
clash_get_current_war |
Current regular war with full per-attack data |
clash_get_cwl_group |
Current CWL group + war tags per round |
clash_get_cwl_war |
Individual CWL war by tag |
clash_get_player |
Player profile (heroes, troops, war stars, current clan) |
clash_api_get |
Generic GET passthrough for any endpoint not wrapped above (gold pass, capital raids, leagues, locations, clan search, etc.) |
Workflow tools (the leverage):
| Tool | Purpose |
|---|---|
clash_grade_war |
Apply your rubric to a war's attacks → per-player grades, scores, rule violations |
clash_war_report |
Markdown post-mortem answering: who attacked, who followed rules, performance leaderboard, smart-attack honor roll, and missed opportunities |
clash_missed_opportunities |
For each attack, check if a weaker undefeated base was sitting there at the time. The deeper version of the smart-attack rule. |
clash_carry_forward_recommendation |
Analyze all CWL rounds → keep / review / bench list for next season |
clash_promotion_candidates |
Screen Members against your elder bar (donation ratio, hero progress, war contribution) |
clash_snapshot_war |
Save current war (regular + CWL rounds) to local disk. Idempotent. Designed to run on a schedule. |
clash_snapshot_status |
Show what's in the snapshot store + identify gaps vs the warlog |
clash_player_war_history |
Pull a player's per-attack history across stored snapshots (attendance %, avg stars, trends) |
Getting started
1. Get a free Clash of Clans API token
This MCP wraps the official Supercell API, which is free but requires a token tied to your IP address.
-
Click Login and sign in with your Supercell ID (the same one tied to your in-game account — you'll get an email code).
-
Once logged in, go to My Account → Create New Key.
-
Find your current public IP — easiest:
curl ifconfig.me -
Fill in the form:
- Key Name: anything (e.g.
coc-mcp) - Description: anything (e.g.
MCP server for clan analysis) - Allowed IP Addresses: paste the IP you got from
curl ifconfig.me
- Key Name: anything (e.g.
-
Click Create Key. Copy the long JWT token that appears (starts with
eyJ...).🛡️ The token is IP-locked. If your IP changes (new ISP, VPN, coffee shop), regenerate the key with the new IP, OR run the MCP from a host with a fixed IP (a small VPS, a home server, etc.).
2. Find your clan tag
In-game, tap your clan name → look for the tag like #YV9JRULU. Note: clan tags use uppercase letters and digits only (no O, only 0).
3. Clone and install
git clone https://github.com/justinritchie/clash-of-clans-mcp.git
cd clash-of-clans-mcp
pip install -r requirements.txt
4. Configure your token
cp .env.example .env
Edit .env:
COC_API_TOKEN=eyJ0eXAiOiJKV1Q... # <-- paste your token
COC_DEFAULT_CLAN_TAG=#YV9JRULU # <-- your clan tag
The .env file is .gitignored, so your token won't accidentally get committed.
5. Smoke test
python coc_test.py
You should see green checkmarks for every endpoint:
✅ GET /clans/<tag>
Clan: Broken Arrow (level 27) — 44 members, 793W
✅ GET /clans/<tag>/members
✅ GET /clans/<tag>/warlog
✅ GET /clans/<tag>/currentwar/leaguegroup
✅ GET /clans/<tag>/currentwar
If you get ❌ HTTP 403, your IP doesn't match the token's whitelisted IP. Regenerate the token with your current IP (curl ifconfig.me).
If the CWL leaguegroup returns 404, that's fine — CWL only runs the first ~10 days of each month.
6. Run the unit tests
pip install pytest
pytest tests/ -v
Eight tests should pass. They run against fixture JSON, no API calls.
7. Try the war report CLI
python run_war_report.py
This dumps the full markdown report for your most recent war (whatever state it's in).
Wire it into Claude Desktop
Option A — Use the included installer (recommended)
python install_to_claude_desktop.py
This:
- Backs up your existing
claude_desktop_config.json(timestamped) - Appends a
cocentry without touching any other MCPs - Pulls the token from your
.env - Idempotent — re-running just updates the entry
Then restart Claude Desktop.
Option B — Manual edit
Add this to ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or %APPDATA%/Claude/claude_desktop_config.json (Windows):
{
"mcpServers": {
"coc": {
"command": "python3",
"args": ["/absolute/path/to/coc-mcp/coc_mcp_server.py"],
"env": {
"COC_API_TOKEN": "eyJ0eXAi...",
"COC_DEFAULT_CLAN_TAG": "#YOURTAG"
}
}
}
}
Restart Claude Desktop. The coc_* tools will appear.
Cowork or other MCP clients
The MCP runs over stdio. Anything that supports stdio MCP servers will work — point it at coc_mcp_server.py with the env vars set.
Customizing the rubric
The default rules live in config/rubric.default.json:
{
"war_attack_rubric": {
"first_attack": {
"acceptable_offsets": [0, 1], // mirror or one down
"min_stars_for_pass": 2
},
"second_attack": {
"preferred_offsets_min": 1, // must attack lower than your rank
"must_3_star": true // smart 3-star expected
},
"missed_attack_penalty": -10
},
"elder_promotion_rubric": {
"min_donation_ratio": 0.4,
"min_hero_progress_pct": 80,
"min_war_score_last_n": 50
},
"carry_forward_rubric": {
"min_attacks_used_pct": 100,
"min_avg_stars": 2.0,
"min_avg_destruction": 75
}
}
Two ways to customize:
-
Permanently — copy to
config/rubric.local.json(gitignored), edit, then point to it:COC_RUBRIC_PATH=./config/rubric.local.json -
Per call — when invoking
clash_grade_warorclash_war_report, passrubric_overrideswith just the fields you want to change. They're deep-merged onto the loaded rubric.Example: relax the smart-3⭐ rule for one analysis:
{ "war_attack_rubric": { "second_attack": { "must_3_star": false } } }
Limitations (be aware before you start)
- Per-attack data is only available for the current war (state
inWarorwarEnded). The/warlogendpoint returns summary only (result, totals — no per-attack data). To analyze past wars over time, snapshot/currentwarto disk while state iswarEnded. (Future enhancement: aclash_snapshot_wartool.) - The API is read-only. You cannot promote, kick, accept join requests, edit clan settings, send messages, or change war state. Supercell intentionally doesn't expose write endpoints. This is a diagnostic / advisory tool — it tells you who deserves the action; you take it in-game.
- No clan chat access. Supercell doesn't expose chat. Use a Discord bridge bot if you need chat-based workflows.
- API token is IP-whitelisted. Token breaks if your IP changes. Regenerate or use a fixed-IP host.
- Tenure / "days in clan" is not in the official API. That data lives on third-party sites like clashofstats.com. Future enhancement could scrape it; for now, the elder rubric uses war contribution and donation ratio as proxies.
Project layout
coc-mcp/
├── coc_mcp_server.py # FastMCP entry — tool registrations
├── coc_mcp/
│ ├── client.py # Async COC API client
│ ├── grading.py # Rubric engine
│ ├── reporting.py # Markdown generators
│ └── config.py # Token + rubric loading
├── config/
│ └── rubric.default.json # Default rubric (edit freely)
├── tests/
│ ├── fixtures/ # Sample war JSON for tests
│ └── test_grading.py # Unit tests
├── coc_test.py # Live smoke-test script
├── run_war_report.py # CLI: dump war report for current war
├── install_to_claude_desktop.py # Safe installer (backs up first)
└── mcp_config_example.json # Drop-in for claude_desktop_config.json
Contributing
PRs welcome — especially:
- More workflow tools for common leadership tasks (war post-mortem trends, donation watch, hero progression alerts)
- Tenure scraping from clashofstats.com or similar (worth its own MCP)
- Snapshot store for historical war analysis (sqlite or jsonl)
- War prediction (given roster + opponent, what's the expected outcome?)
- More language SDKs (this is Python; a TypeScript port would help)
If you build something off this, drop a link — would love to see what other clans are doing.
License
MIT — see LICENSE.
Acknowledgements
This product uses data from the Clash of Clans API but is not affiliated with, endorsed, sponsored, or specifically approved by Supercell. Supercell is not responsible for it. For more information see Supercell's Fan Content Policy: supercell.com/fan-content-policy.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.