Koko Credit card assistant

MCP server to access Koko's credit card workflows to search, compare and optimize credit cards

A hosted MCP server for AI-powered credit card advice. Search 100+ US credit cards, compare options side by side, analyze portfolio health, and get personalized recommendations — all through the Model Context Protocol.

Server URL:https://kokofinance.net/mcp/

No API key required. No local installation needed. Just connect your MCP client.

Add to~/Library/Application Support/Claude/claude_desktop_config.json(macOS) or%APPDATA%\Claude\claude_desktop_config.json(Windows):

{ "mcpServers": { "koko-finance": { "type": "http", "url": "https://kokofinance.net/mcp/" } } }

claude mcp add koko-finance --transport http https://kokofinance.net/mcp/

Or copy the.mcp.jsonfrom this repo into your project.

Go toSettings > MCPand add a new server:

- Name:koko-finance
- Type:HTTP
- URL:https://kokofinance.net/mcp/

{ "mcpServers": { "koko-finance": { "type": "http", "url": "https://kokofinance.net/mcp/" } } }

Several tools accept optional parameters that improve result relevance. All are optional — when omitted, the server uses sensible defaults (e.g., national average spending, no issuer bias).

Credit score range that affects card eligibility and approval likelihood.

Used by:search_credit_cards,compare_cards,optimize_portfolio,recommend_card_for_category,check_card_renewal

The user's main objective for their credit card strategy.

Used by:compare_cards,optimize_portfolio,recommend_card_for_category,check_card_renewal

A list of issuer preference objects that express affinity or aversion toward specific card issuers. Each entry has anissuername and aweightbetween -1.0 (strong aversion) and 1.0 (strong preference).

[ {"issuer": "Chase", "weight": 0.8}, {"issuer": "American Express", "weight": 0.5}, {"issuer": "Capital One", "weight": -0.5} ]

Used by:compare_cards,optimize_portfolio,recommend_card_for_category

A dictionary of monthly spending amounts by category, used to calculate personalized reward values instead of national averages.

{"dining": 500, "groceries": 800, "travel": 300, "gas": 150, "general": 2000}

Used by:compare_cards,optimize_portfolio,recommend_card_for_category,check_card_renewal,calculate_card_value

Override the default net-value thresholds (in dollars) for KEEP/OPTIMIZE/CANCEL verdicts.

- cancel_threshold— Cards below this → CANCEL (default: -50)
- keep_threshold— Cards above this → KEEP (default: 50)

renewal_cancel_threshold/renewal_keep_threshold

Same concept applied to a single card's renewal decision.

Date the user acquired the card (YYYY-MM-DD). Informs tenure-based renewal analysis.

Calculation tools include metadata fields in responses:

- spending_source—"user_provided"or"national_averages". Indicates whether calculations used user-supplied spending or BLS national averages.
- data_source— Indicates where card data was sourced (e.g.,"koko_db").
- rewards_by_category— Per-category reward breakdown included in compare, portfolio, and value tools.
- points_program_key/portal_cpp— Points program identifier and portal cents-per-point valuation, included on all card data.

Pre-built conversation starters that guide the AI through structured workflows:

Theoptimize_portfolioandcheck_card_renewaltools accept abenefit_selectionsparameter — a list of individual benefit keys the user actually uses. Selected benefits count at 100% utilization; unselected benefits count at 0%.

benefit_selections: ["uber", "airline_fee", "dining", "admirals_club"]

This gives accurate net-value calculations instead of the default 50% utilization estimate.

Discover valid keysby calling the REST endpointGET https://kokofinance.net/api/v1/benefit-categories(no auth required), or see thedeveloper docs.

- KoKo Finance— Main site
- Developer Documentation— Full API and tool reference
- MCP Endpoint— Direct server URL

- Claude Code Plugin— Adds the MCP server connection plus an Agent Skill that teaches Claude structured credit card workflows

Full privacy policy:https://kokofinance.net/privacy.html

Power up your AI agents with 31 production-ready tools. Features local-first Python analysis, real-time SearXNG search, and a secure local proxy to sanitize prompts. Built for developers who value performance and privacy. Install once, use everywhere.