poe_mcp_suite
AGENTS.md9 starsChanged 3 months agoMITView on GitHub
This file is the vendor-neutral entry point for any MCP-compatible AI agent (Cursor, Windsurf, Cline, Codex, Zed, Gemini CLI, custom agents, …). It routes you to the shared, plain-Markdown resources in this repo. If you are Claude Code, read CLAUDE.md instead/as well — it adds Claude-specific mechanics (auto-triggering skills, a SessionStart hook, memory) layered on top of everything below. A collection of MCP servers that give an agent deep, live integration with Path of Exile — build theory-crafting and passive-tree…
In this repository23 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- atlas-planningSkill
- build-comparisonSkill
- build-designSkill
- build-optimization-simSkill
- build-shortlistSkill
- character-levelingSkill
- community-surveySkill
- crafting-lookupSkill
- crafting-optimizationSkill
- dps-analysisSkill
- freshness-checkSkill
- gear-crafting-auditSkill
- gear-shoppingSkill
- gem-socket-analysisSkill
- guide-analysisSkill
- knowledge-frameworksSkill
- league-transitionSkill
- snapshot-driven-build-guideSkill
- spectre-analysisSkill
- stash-scanningSkill
- tree-analysisSkill
- verify-installSkill
The file
Files
poe_mcp_suite — Guide for AI Agents
This file is the vendor-neutral entry point for any MCP-compatible AI agent
(Cursor, Windsurf, Cline, Codex, Zed, Gemini CLI, custom agents, …). It routes
you to the shared, plain-Markdown resources in this repo. If you are Claude
Code, read CLAUDE.md instead/as well — it adds Claude-specific
mechanics (auto-triggering skills, a SessionStart hook, memory) layered on top of
everything below.
What this is
A collection of MCP servers that give an agent deep, live integration with Path
of Exile — build theory-crafting and passive-tree simulation via Path of
Building, trade/economy lookups, wiki and crafting data, and loot filters. Each
server runs independently; the tools are agent-agnostic by construction (standard
MCP). See README.md for the server/tool overview and
INSTALL.md for setup (submodules, dependencies, .mcp.json).
The current league name is in .mcp.json as POE_LEAGUE; pass it explicitly on
tool calls rather than trusting auto-detection when leagues run concurrently.
How to work in this repo
The suite uses a lightweight workflow framework. It is described once, in
playbooks/README.md — the shared meta-framework
(context management, league pre-flight, trust hierarchy, narration norms). Read
it before a substantial task, then the specific playbook that matches. The
playbooks are ordinary Markdown — any agent can read and follow them directly,
even without skill/rule auto-loading. Key ideas:
- Cursory-vs-detailed gate — decide up front whether a request is a quick
lookup (answer directly) or a multi-source/multi-stage analysis (engage the
matching playbook and pause for approval before pulling data).
playbooks/README.md§1. - Freshness discipline — the game changes every patch, and model training
data lags. For any patch-specific fact (mods, node stats, unique stats,
prices, league specifics), prefer fetching authoritative data via the tools
over answering from memory. Model-aware staleness index:
reference_data/freshness_index.md. - Cached game data lives in
reference_data/(gitignored but itsREADME.mdis committed — read it on a fresh clone). Check it before re-fetching; respect thefetched:/patch:/league:frontmatter. - Per-character analyses live in
character_data/(see itsREADME.md). Read a character's file before working on it; append findings as you go.
Non-negotiables
- Legal boundaries — before extracting, modifying, or redistributing any
Path of Exile game data, read
legal_considerations.md. Stay inside the conservative redistribution boundary it documents; if unsure, ask the user. - GGG trade API — the official trade API carries Terms-of-Service risk.
Brief the user and get explicit consent before using trade tools (see
legal_considerations.md→ "GGG Terms of Service"). - Secrets —
.mcp.jsonandPOE_SESSION_IDare sensitive (session cookie / account access). Never log or commit them.
Contributing back
New or improved playbooks, reference data, and pitfalls are valuable to everyone
— see playbooks/README.md for format conventions and
submission guidelines.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

