Agentcity Robinhood Chain Research

What Agentcity Robinhood Chain Research MCP server does

Agentcity Robinhood Chain Research MCP server provides read-only token research for Robinhood Chain through GMGN data. It is intended for agents that need to find tokens, select a contract address, gather supporting market and holder information, and produce a research-oriented report. The server does not place trades or expose swap, order, wallet, signing, or private-key functionality.

Launch discovery supports new creations, near-completion launches, and completed launches. Results can be limited by liquidity and filtered to Pons, Trench, or all launchpads indexed by GMGN. A separate screening tool applies GMGN server-side quality profiles named safe, smart-money, and strict.

Token research combines information about the token, security, liquidity pools, holders, and top traders. Research and comparison responses also include a deterministic research score. The score is a prioritization aid rather than a profit estimate, investment recommendation, or trading instruction.

How it works

The typical workflow begins with launch lists, rankings, signals, smart-money activity, or a name-based search. Search returns candidate contracts, after which the agent should use a 20-byte EVM address rather than a symbol for detailed research. The agent can then request a full packet, inspect OHLCV candles, or compare two to five addresses.

Agentcity Robinhood Chain Research MCP server is available through the hosted streamable HTTP endpoint https://gmgn.agentcity.lol/mcp. The README also documents local stdio usage. A local deployment calls the Node.js entry point and uses the GMGN CLI and API key to retrieve data.

The score uses five weighted areas: data completeness, liquidity depth, holder distribution, contract safety, and activity quality. Hard flags include wash trading, a rug ratio above 0.3, or top-10 holder concentration above 50%. These flags can result in a rejection from the research queue regardless of the numeric score.

Setup and configuration

The documented runtime requirement is Node.js 18 or newer, gmgn-cli version 1.6.0 or newer, and a GMGN API key. The hosted endpoint can be added to Claude Code, Claude Desktop, Cursor, or VS Code using each client's MCP configuration. Local stdio configuration sets MCP_TRANSPORT to stdio and supplies GMGN_API_KEY.

No GMGN_PRIVATE_KEY is needed. The server is designed for market-data research and does not require credentials for trading or wallet control.

Tools and capabilities

• List launchpad tokens by stage, liquidity threshold, and launchpad.
• Screen launches with quality profiles and risk-related thresholds.
• Search by token name, symbol, or address to find candidate contracts.
• Retrieve trending rankings across intervals from one minute through 24 hours.
• Fetch full token research, holder and trader data, and a derived score.
• Read OHLCV candles at resolutions from 30 seconds through one day.
• Compare two to five Robinhood Chain token addresses.
• Query GMGN-tagged smart-money or KOL trades.
• Retrieve GMGN market signals with market-cap and signal-type filters.
• Build a daily brief containing new launches and 24-hour trending data, with optional smart-money activity.

Limitations and notes

Agentcity Robinhood Chain Research MCP server reads market data only. Its outputs are research inputs for human decisions, not trade instructions. Contract addresses must use the 0x prefix followed by 40 hexadecimal characters; symbols are intentionally rejected for detailed address-based operations because they are not unique.

Tool limits include up to 80 launch results, 100 trending results, 200 smart-money trades, and 100 holders or traders. Signal types accepted by the server are 1–13 and 17–21; types 14–16 are blocked because GMGN rejects them. The README identifies the score methodology as agentcity-gmgn-research-v1, and its example values are illustrative rather than live results.