Loading ToolBench
MCP server for Gainium — manage trading bots, deals, and balances via AI assistants
Gainium MCP defines 17 tools with consistent verb_noun naming (list_bots, get_bot, create_bot, manage_bot, etc.). All tools have basic descriptions and input schemas with enum constraints on bot/deal types. However, descriptions are generic and lack WHEN-to-use guidance, parameter descriptions are minimal, and output schemas are not documented. The server follows basic naming conventions well but falls short on description depth, parameter guidance, and schema completeness, typical of community servers landing in the 50-69 (C/D) range.
Get information about a backtest run
Clone an existing bot to create a new one
Create a new trading bot with specified settings
Create a new deal
Discover available exchanges, trading pairs, and asset information
Get user account information and balance overview
Get a single bot by its ID and type
Output schemas not documented. Tool descriptions do not explain what fields are returned or their types. LLMs cannot plan downstream tool calls or extract specific data.
Parameter descriptions are minimal or missing context. e.g., 'settings' in create_bot and update_bot is described only as 'Configuration settings for the bot', what fields does settings accept? What are the constraints?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Get a single deal by its ID and type
Get market screener data for cryptocurrency pairs
List all bots of a given type (dca, grid, combo, terminal)
List all deals of a specified type
Manage bot state: start, stop, pause, resume, or change trading pairs
Manage deal state: close, add funds, sell at market, or manage take-profit/stop-loss
Get or set global variables for trading automation
Run a backtest simulation for a bot configuration
Update settings of an existing bot (DCA and Combo only)
Update settings of an existing deal
Tool descriptions lack WHEN-to-use guidance and prerequisites. e.g., 'Get a single bot by its ID and type' does not explain when to call this vs. list_bots, or what happens if the bot does not exist.
run_backtest has an empty input schema (payload object with no field constraints). LLMs cannot validate the payload structure, they will hallucinate fields and waste tokens on retries.
manage_bot and manage_deal have complex, optional parameters that depend on the action value (e.g., closeType only applies to DCA stop). These dependencies are hinted at in descriptions but not formally encoded (no JSON Schema dependencies or conditional logic). LLMs will pass irrelevant params.
No error handling guidance documented. Tool descriptions do not explain what happens on failure (e.g., 'If bot not found, returns 404 with available bot IDs') or whether calls are retryable. Agents have no recovery path.
Destructive operations (manage_bot stop/pause, manage_deal close) lack confirmation or dry-run capability. No warning in descriptions about irreversible consequences.
discover tool name is generic. 'Discover' alone does not clarify whether it discovers exchanges, trading pairs, asset info, or something else. Better name: discover_trading_venues or discover_market_data.