Ashen Throne MCP Guide: Play with ChatGPT or Claude
Ashen Throne MCP connects an AI assistant to your live game account. Instead of copying your troop counts, timers, resources, battle reports, and map situation into a chat, you can ask ChatGPT, Claude, or Codex to inspect them through Ashen Throne's tools. The assistant can explain your position, find a bottleneck, simulate an extraction plan, and, if you grant acting permission, carry out individual approved steps.
The best way to begin is read-only. Ask for an account overview, request a plan, inspect the costs and risks, then decide whether to authorize actions. Ashen Throne keeps acting permission separate, requires confirmation when something is at stake, and gives you an in-game STOP control.
If you still need an account, start the Ashen Throne open beta. If you are ready to connect, keep the official MCP setup page open beside this guide.
What MCP changes for an Ashen Throne player
Model Context Protocol is an open standard that lets AI applications connect to external data and tools. In Ashen Throne, that connection is scoped to the player who signs in and authorizes it.
A normal chatbot knows only what you type. A connected assistant can ask Ashen Throne for the current state behind your question:
| Your question | What the connection adds |
|---|---|
| What should I do next? | Your actual facilities, resources, objectives, active squads, research, and timers |
| Who can lead this run? | Every owned commander, current squad, equipment, Leadership, injuries, and deployment state |
| Can I take this site? | The site's wave sequence, troops you genuinely have free, deploy cost, and engine-simulated matchups |
| What happened in that battle? | Your battle report and, with VIP Growth Status, a re-fight through the real combat engine |
| Is anything near my Crawler? | Only the Crawlers, sites, Hunters, marches, and other objects your own map visibility permits |
| How does this rule work? | Game knowledge generated from Ashen Throne's current configuration instead of a remembered guide |
That makes the assistant useful for synthesis. It can connect facts spread across the Barracks, map, Objectives panel, battle reports, and running timers without asking you to transcribe each screen.
Connect Claude, ChatGPT, or Codex
The server address is:
https://mcp.ashenthr.one/mcp
Use the connection method supported by your AI client. Interface labels can change, so the Ashen Throne MCP page remains the canonical setup reference.
Claude
Claude supports remote MCP servers as custom connectors. In Claude, open Settings, choose Customize, open Connectors, add a custom connector, and paste the Ashen Throne MCP address. Team and Enterprise workspaces may require an owner to add the connector before members connect it. Claude's official custom connector guide has the current account and plan-specific steps.
After adding the connector, choose Connect. Ashen Throne opens its own authorization screen. Sign in to the game account you want the assistant to use, review the requested permissions, approve them, and return to Claude. Enable the connector in the conversation where you want it available.
ChatGPT
ChatGPT web receives remote MCP tools through apps or plugins. Use the Ashen Throne connection when it is available to your account, then complete the same Ashen Throne sign-in and permission review. Availability can vary during directory review, so this guide does not assume that every ChatGPT account can already find the listing. OpenAI's current MCP guide explains how ChatGPT exposes remote MCP tools.
Codex
Codex can connect directly to a remote MCP server. In the Codex desktop app, open Settings, add an MCP server, enter the Ashen Throne address, and authenticate. From the Codex command line, the equivalent flow is:
codex mcp add ashen-throne --url https://mcp.ashenthr.one/mcp
codex mcp login ashen-throne
The browser authorization belongs to Ashen Throne. Your AI client receives OAuth access, not your game password.
Your first useful MCP session
Do not start by telling the assistant to automate everything. Start with one decision and make it show its work.
1. Establish the live position
Ask:
Give me my Ashen Throne account overview. Summarize my Crawler zone and position, free resources, active commanders, idle troops, current objectives, deployed squads, and anything finishing in the next hour. Do not act.
This creates a shared baseline. If a resource is capped, a commander is injured, or a squad is already deployed, later advice should account for it.
2. Find the immediate bottleneck
Ask:
Based only on the live account and current game rules, what is the single best progression action I can take next? Give me the reason, prerequisites, opportunity cost, and anything I should wait for. Do not act.
This is more useful than a generic build order because it starts from what your account can do now. New players can compare the answer with the first 24 hours guide.
3. Request an extraction plan
Ask:
Plan the best reachable Frontier extraction for me. Stay out of the Wastes. Prefer an idle commander, never change the Crawler defender, and show the site, expected wave classes, recommended squad, deploy bill, travel time, and simulated matchup. Do not act.
The planner first grades the commander's current squad against the complete wave sequence. If that squad is inadequate, it searches formations and other idle commanders using only troops that are actually free. The result is a proposal, not a march order.
4. Inspect what will change
Before granting action, check four things:
- which commander and recipe slot will change;
- which troops leave reserve;
- what Metal, Energy, and Food the deployment spends;
- whether movement, site occupation, or combat exposes anything you are not willing to risk.
If the recommendation is unclear, ask the assistant to compare two sites or explain the weakest simulated wave. You do not need acting permission to improve a plan.
5. Approve one operation at a time
If you want the assistant to execute the plan, grant the separate acting permission and say exactly what it may do:
Carry out this Frontier extraction plan one step at a time. Tell me before any step that needs confirmation. Do not enter the Wastes, spend a Temporal Catalyst, recall another squad, or change the Crawler defender.
The assistant may activate a commander, set a squad, leave a city, move or boost the Crawler, deploy, and recall. Actions with a meaningful cost or risk refuse to run until the player has agreed and the tool is called with confirmation. The server uses the same game routes and constraints as the interface, including Leadership, range, transit, squad state, deploy costs, and daily extraction limits.
6. Watch and interrupt
An act-capable connection displays an AI ASSISTING chip in the game. Accepted actions are narrated there. Press STOP and the next attempted act is refused, even when the assistant is midway through a task.
The assistant can wait for a Crawler arrival, extraction wave, return, recall, or nearby visibility change and then refresh the relevant status. This lets one conversation follow a run without creating standing orders on the server. The AI still takes one tool call at a time.
Prompt recipes worth saving
Use prompts that specify the decision, boundaries, and desired evidence.
Daily planning
Review my live account and give me a three-step plan for today. Consider capped resources, idle queues, wounded recovery, objectives, and active marches. Quote exact numbers from the tools. Do not act.
Timer check
List everything finishing in the next two hours, ordered by time. Tell me which completion unlocks another useful action and whether any free boost is currently available. Do not act.
Army audit
Audit my active commanders and squads. Find stranded reserve troops, Leadership waste, injured commanders, and poor class matchups. Recommend changes but do not apply them.
Extraction comparison
Compare the two best Frontier extraction choices I can reach. Use the full wave sequence, troops actually free, deploy cost, travel time, and expected resource value. Recommend one and explain the tradeoff. Do not act.
Battle review
Show my five most recent battles. For the latest loss, separate what the report proves from what requires simulation. If battle analysis is available, use it and quote its measured results exactly.
Situational awareness
What can my account currently see around the Crawler? Summarize nearby players, occupied extraction sites, Hunters, and incoming or outgoing marches. Do not infer anything hidden by fog.
How Ashen Throne keeps the player in control
The control model has several layers:
- OAuth authorization. You sign in on Ashen Throne's page. The client does not receive your game password.
- Separate read and act access. You can use account-aware advice without allowing the assistant to change the game.
- Explicit confirmation. Leaving a city, moving the Crawler, spending a Temporal Catalyst, deploying troops, and recalling a run require agreement when the tool identifies something at stake.
- Game-route enforcement. The assistant cannot bypass the rules that apply to your own taps.
- Hard behavioral boundaries. It does not touch the Crawler defender, enter the Wastes unless you asked, or spend ASH or $ato.
- Visible interruption. Press STOP in the AI ASSISTING chip to pause acts. Resume or revoke a connection from Profile > Settings > Account.
Revocation applies to the connection itself. Use it when you no longer want that client to read or act on the account.
Free and VIP limits
MCP is available during the free-to-play open beta. Tool calls are limited per player and reset at 00:00 UTC.
| Access | Read calls per day | Act calls per day | Knowledge and analysis |
|---|---|---|---|
| Free | 100 | 60 | Core game knowledge |
| VIP Growth Status | 400 | 200 | Full archive plus engine-backed battle analysis |
The limits are safeguards, not a recommended target. A focused overview and planning session should not need hundreds of calls. Waiting for a tracked game-state change is exempt from the normal read-call budget.
What Ashen Throne MCP does not do
It is not a guaranteed-win button. Simulations improve the evidence behind a decision, but battles and other players still create uncertainty. It is not a hidden map: surroundings are limited to what your account can see. It is not an unattended server bot: there are no standing orders, consequential steps require player agreement, and the player can stop or revoke access.
Most importantly, MCP does not replace the strategic choice. It reduces the work of gathering current facts and carrying out an approved sequence. You still decide what is worth building, where your Crawler should go, and which troops you are prepared to risk.
Connect your strategist
Start with read-only access, ask for your live account overview, and make the assistant earn your trust with one useful plan. Grant acting permission only when you want it to execute a specific operation.