About cluster-mcp
cluster-mcp is a self-hostable MCP server and skills platform for AI financial agents on Robinhood Chain. It provides tools for tokenized stocks and crypto, swaps, market data, finance, and persistent memory, and can target a cluster deployment.
Key features
- 0135 MCP tools for agent workflows
- 02Tokenized stock and crypto capabilities
- 03Uniswap v3 swap workflow
- 04Market data and portfolio tools
- 05Persistent memory with Hindsight option
Use cases
- 01Query quotes and market information
- 02Execute a swap with wallet signing
- 03Review portfolios and payouts
- 04Store agent observations and memory
MCP support
The MCP server exposes 35 tools across finance, trading, memory, and related skills. Clients connect it to a cluster deployment using the API URL and credentials.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.clusteragent.dev/mcpclaude mcp add --transport http dev-clusteragent-cluster https://mcp.clusteragent.dev/mcp{"mcpServers":{"dev-clusteragent-cluster":{"type":"http","url":"https://mcp.clusteragent.dev/mcp"}}}Open in Cursor{"servers":{"dev-clusteragent-cluster":{"type":"http","url":"https://mcp.clusteragent.dev/mcp"}}}Open in VS Code{"mcpServers":{"dev-clusteragent-cluster":{"type":"http","url":"https://mcp.clusteragent.dev/mcp"}}}{"mcpServers":{"dev-clusteragent-cluster":{"serverUrl":"https://mcp.clusteragent.dev/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://mcp.clusteragent.dev/mcpAuth requiredTools
Details
From the project READMEcluster
The runtime layer for agentic finance on Robinhood Chain. Self-hostable index of AI financial agents — 192 tokenized stocks + native crypto, real Uniswap v3 swaps (non-custodial), persistent memory, and a 24-model LLM gateway.
npx @clusteragent/cluster-mcp
npx skills add clusteragent/cluster # any skills-CLI agent
pip install cluster-agent # Python SDK
Repo Layout
├── SKILL.md main installer skill (any agent)
├── index.mjs MCP server — 35 tools (@clusteragent/cluster-mcp)
├── skills/
│ ├── swap/ dedicated swap execution: safety scan → quote →
│ │ calldata → sign → receipt proof, DCA/bracket patterns
│ ├── trading/ quotes, market data, chart (4663)
│ ├── memory/ built-in memory + Hindsight backend
│ ├── finance/ portfolio, payouts, keys & metering
│ ├── crypto-intel/ wallet forensics, pool forensics, scout swarm
│ ├── research/ agent roster + OpenBB integration
│ ├── llm-gateway/ 24 models, thinking mode, credit accounting
│ ├── portfolio/ wallet valuation, DCA/TP-SL previews
│ ├── social/ X links + tip intents
│ └── monitor/ price/wallet/cycle/gas watch loops
├── python/ cluster-agent PyPI SDK
└── docs/self-hosting.md deployment guide
11 skills total — 1 installer + 10 domain sub-skills.
The cluster backend, frontend and keeper bot are not in this repo — the bot signs real treasury transactions, so the operational stack stays private. This repo is the skill/MCP platform: point it at any cluster deployment.
Quick Start
Agent users — see SKILL.md. MCP config:
{ "mcpServers": { "cluster": {
"command": "npx", "args": ["-y", "@clusteragent/cluster-mcp"],
"env": { "CLUSTER_API_URL": "https://clusteragent.dev", "CLUSTER_API_KEY": "clst_...", "CLUSTER_WALLET": "0x..." }
}}}
Self-hosters — see docs/self-hosting.md. Docker images + a deployment guide; SQLite default, zero external services except one LLM key for chat.
Python — see python/:
from cluster import Cluster
c = Cluster(api_url="https://your-host", api_key="clst_...")
c.quotes(["NVDA", "PONS"]); c.payout_basket(); c.chat("hi", thinking=True)
The Mechanism
fees in ──▶ vault ──▶ keeper bot buys the 19-name basket ──▶ $CLST holders paid pro-rata
Holding is the position. Every cycle is public: GET /api/distributions.
Integrations
| Project | Role | Guide |
|---|---|---|
| Hindsight | learning memory backend (observations, mental models) | skills/memory/ |
| OpenBB | institutional market data (fundamentals, macro) via MCP/Python | skills/research/ |
| Maybe | self-hosted personal finance UI (AGPLv3 — rename your fork) | skills/finance/ |
| OpenCatz | multi-agent scout swarm pattern on 4663 | skills/crypto-intel/ |
| Uniswap AI | skills architecture this repo follows | structure |
The Agents
Trading Relay · Pivot — Research Scout · Sifter · Quill — Analysis Argus · Prism · Census — Finance Ledger · Remit · Margin — Memory Memoria · Echo — Crypto Nexus · Vault · Oracle
Honesty by Construction
- Pre-launch payout data = honest zeros, never fabricated
- Swaps: API quotes, your wallet signs — no private keys server-side
- Keys: SHA-256 at rest, shown once
- Quote = prediction; receipt = proof (parse the Transfer logs)
License
MIT. Backend and skills are original work. The Maybe Finance fork note applies if
you self-host that integration (AGPLv3, trademark rules in skills/finance/).
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.