mtgmon Mortgage Analytics
Live · Public · Read-only

Mortgage analytics
for AI assistants.

Give ChatGPT, Codex, and other MCP clients structured access to mtgmon market, lender, GSE performance, and servicing data—with source scope, units, as-of dates, and caveats attached.

Connect a client View metadata
Streamable HTTP endpoint https://mtgmon.com/mcp
Authentication
None
Access
Read-only
Rate limit
60 requests/minute
Capabilities

Seven focused tools, bounded responses.

The server favors compact, analysis-ready results over raw data dumps. Every analytical response includes provenance and interpretation guardrails.

get_market_snapshot

Primary rates, TBA coupons, Treasury and SOFR benchmarks, and MBS basis attribution.

get_loan_performance

GSE credit, prepayment, delinquency, transitions, modifications, losses, and lock-in.

search_lenders

Find canonical HMDA lender keys by name, charter, state, or volume.

get_lender_profile

Volume, rank, market share, geography, and product mix for one lender group.

compare_lenders

Normalized side-by-side HMDA comparison of two canonical lender groups.

get_servicer_analytics

Agency servicer performance, scorecards, and lender servicing-transfer analysis.

get_dataset_status

Source coverage, freshness, build, and cache metadata.

Connect

Use the public endpoint.

No local server process or API key is required.

ChatGPT

Create a developer-mode app in ChatGPT, use the name mtgmon Mortgage Analytics, and enter the endpoint below as its MCP URL.

https://mtgmon.com/mcp

Codex CLI

Add the hosted server to your MCP configuration from a terminal.

codex mcp add mtgmon --url https://mtgmon.com/mcp

Other MCP clients

Choose Streamable HTTP and use the public URL directly. The server is stateless and does not require custom headers.

https://mtgmon.com/mcp
Trust & operations

Observable without collecting prompts.

Read-only by contract. All tools carry non-destructive, idempotent annotations and cannot change mortgage data.

Operational monitoring. Aggregate request counts, latency, errors, availability, and tool usage are monitored. Tool arguments and response payloads are not stored in metrics.

Source-aware output. Results distinguish HMDA originations from GSE servicing performance and preserve weighting, units, dates, and caveats.