Connect your AI agent to Kova and evaluate portfolios against investment strategies. The fastest way to get started is the Claude.ai connector -- connect in seconds, no install needed.
Kova exposes an MCP server that any compatible AI agent can discover and use automatically.
Connect Kova directly to Claude.ai -- no CLI, no install. Once connected, Kova tools are available in every conversation.
Prerequisites: A Kova account. Sign up if you don't have one.
Kova as the name and https://kovatools.com/mcp as the URLKova now appears alongside your other connectors. In any Claude.ai conversation, just ask naturally — no special syntax needed. Claude discovers and calls Kova tools automatically.
Example prompts:
Troubleshooting: If the connector shows an error, make sure you're logged in to kovatools.com and try re-authorizing from the Integrations page.
For Claude Code, Cursor, Claude Desktop, or any MCP-compatible client. Install the CLI, authenticate, then register Kova with your agent.
Homebrew (macOS / Linux):
brew install kovatools/kova/kova
Or download the latest binary from GitHub Releases.
Verify the installation:
kova --version
Browser login (recommended):
kova auth login
Opens your browser for secure authentication. Your token is stored locally at ~/.config/kova/config.json.
Manual token: Generate a token in Settings, then:
kova auth login --token kova_tk_xxxxxxxxxxxxxxxxxxxx
kova agent install
This detects Claude Code and Claude Desktop on your system and registers Kova as an MCP tool provider. Restart your client after installation.
That's it -- Kova is now available inside your coding agent. Just ask it to list your strategies, evaluate a portfolio, or create a new strategy. No special syntax, no imports -- your agent discovers Kova automatically.
For other MCP-compatible clients, add this to your MCP configuration:
{
"mcpServers": {
"kova": {
"command": "kova",
"args": ["mcp", "serve"]
}
}
}
Config file locations:
| Client | Config File |
|---|---|
| Claude Code | ~/.claude.json |
| Claude Desktop | ~/Library/Application Support/Claude/claude_desktop_config.json |
Once connected, your AI agent has access to these tools:
| Tool | Description |
|---|---|
kova_list_strategies |
List all strategies with alignment scores and labels |
kova_show_strategy |
Read a strategy's full content and metadata |
kova_create_strategy |
Create a new investment strategy with a name and content |
kova_save_strategy |
Update an existing strategy's name, content, or both |
kova_list_strategy_versions |
List all versions of a strategy |
kova_show_strategy_version |
Read a specific version's content |
kova_restore_strategy_version |
Restore a strategy to a previous version |
kova_evaluate_portfolio |
Submit a portfolio, run evaluation, return the completed brief |
kova_simulate_portfolio |
What-if evaluation against a hypothetical portfolio — no data is saved |
kova_show_brief |
Read a completed brief by ID (or "latest") |
kova_list_briefs |
List all briefs for a strategy |
kova_evaluate_portfolio handles polling internally. The agent submits the portfolio and gets back the completed brief -- no manual polling required.
Example prompts for your agent:
kova strategies list # List all strategies
kova strategies show <uuid> # Show strategy details
kova strategies create --name "Growth" \
--content "40% VOO, 30% QQQ..." # Create a strategy
kova strategies create --file strat.md # Create from file
kova strategies update <uuid> \
--content "updated content" # Update a strategy
kova strategies delete <uuid> # Delete a strategy
kova strategies versions <uuid> # List version history
kova strategies versions <uuid> \
--show 3 # Show version 3 content
kova strategies restore <uuid> 3 # Restore to version 3
kova brief run --strategy <uuid> \
--portfolio portfolio.csv # Generate a brief
kova brief run --strategy <uuid> \
--portfolio "AAPL 100, MSFT 50" # Inline portfolio
kova brief show latest --strategy <uuid> # Most recent brief
All commands support --format json for machine-readable output.
The REST API that powers both the CLI and MCP server is documented at API Docs.