vynn-mcp
# vynn-mcp
mcp-name: io.github.beee003/vynn-mcp
MCP server for [Vynn](https://the-vynn.com) — self-improving AI workflows & backtesting.
## Tools
### Workflows
- `list_workflows` — List your workflows
- `get_workflow` — Get workflow details with steps
- `create_workflow` — Create a new workflow
- `run_workflow` — Execute a workflow
- `get_runs` / `get_run_summary` — View run history
### Self-Improving
- `optimize_prompt` — AI-powered prompt optimization for a step
- `apply_prompt_optimization` — Apply an optimized prompt
- `get_model_recommendation` — Get model swap recommendations
- `set_schedule` / `get_schedule` / `delete_schedule` — Cron scheduling
- `create_trigger` / `list_triggers` — Inbound webhook triggers
### Backtesting
- `backtest` — Run a backtest with natural language or structured strategies
- `batch_backtest` — Parameter sweep across strategy variations
- `optimize_portfolio` — Mean-variance portfolio optimization
### Utilities
- `list_templates` / `clone_template` — Workflow templates
- `list_available_tools` — Tools attachable to workflow steps
- `get_analytics` — Workflow performance analytics
## Setup
```bash
# Install
pip install vynn-mcp
# Or from source
pip install -e /path/to/vynn-mcp
```
Set your API key:
```bash
export VYNN_API_KEY="vynn_free_..."
```
## Usage with Claude Code
Add to your Claude Code MCP config (`~/.claude.json`):
```json
{
"mcpServers": {
"vynn": {
"command": "vynn-mcp",
"env": {
"VYNN_API_KEY": "vynn_free_..."
}
}
}
}
```
## Usage with Cursor
Add to `.cursor/mcp.json`:
```json
{
"mcpServers": {
"vynn": {
"command": "vynn-mcp",
"env": {
"VYNN_API_KEY": "vynn_free_..."
}
}
}
}
```
## Get an API Key
Sign up at [the-vynn.com](https://the-vynn.com) or:
```bash
curl -X POST https://the-vynn.com/v1/signup \
-H "Content-Type: application/json" \
-d '{"email": "you@example.com"}'
```
TDQS
Scored across 21 tools
Most tools target distinct actions or resources (e.g., create_workflow vs list_workflows, backtest vs batch_backtest). Minor overlap between optimize_prompt (generates variants) and apply_prompt_optimization (applies a prompt) but descriptions clarify the difference.
All tool names follow a consistent verb_noun pattern in snake_case (e.g., create_workflow, set_schedule, list_triggers). No mixed conventions or irregular names.
21 tools is above the typical well-scoped range (3-15). While each tool has a clear purpose, the number feels slightly heavy for a single server, especially covering both workflow management and trading optimization.
Workflow management lacks delete_workflow and update_workflow, leaving gaps in lifecycle coverage. Trading tools cover backtesting and optimization but are limited compared to a full trading platform. The set is functional but not fully comprehensive.