Cashfree runs an MCP server at mcp.cashfree.com. It covers the operational side of a payment gateway — orders, payment status, refunds and settlements — which is where support and finance questions actually land.
https://mcp.cashfree.com/mcp
Claude Sonnet 4.5
Chat with 60+ AI models on the same workflow — switch to a different model mid-conversation and re-run the same prompt, or use Compare mode to put several side-by-side and balance quality vs. cost.
Cashfree app credentials. Sandbox and production are separate environments — confirm which one the credential belongs to before an agent touches refunds.
How models use it and what it is built for.
The server exposes Cashfree’s payment APIs as MCP tools. Orders and payments are readable by id or filter, so an agent can answer the question support teams get all day: what happened to this specific transaction, and why did it fail. Refunds can be inspected and, depending on your credential scope, initiated. Settlement data is available, which is the finance-side question — what was actually paid out, when, and which transactions it covered. Payment links and beneficiary details round it out. Because Cashfree operates on Indian payment rails, the failure modes it surfaces are the locally specific ones — UPI collect timeouts, bank-side declines, mandate issues — that a generic payments integration describes poorly. Sandbox and production are separate environments with separate credentials, and pointing an agent at the wrong one is the usual first mistake.
Typical tools an AI model can call. Exact names vary by version.
Copy any of these into MCP Agent Studio after connecting.
Why did order 4821 fail, and was the customer charged?
List refunds initiated this week that have not yet settled.
Which settlement batch included the payments from last Tuesday?
Summarise payment failure reasons for the last thousand transactions.
This is not a single-model product: you get the same MCP connection with 60+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Cashfree
Claude Sonnet 4.5
Payment states are subtle — authorised is not captured, and failed is not always unpaid. Sonnet 4.5 respects that distinction rather than collapsing it into success or failure.
Listing tools proves the server is reachable, not that a model can work with it. Evals go further: they read every tool on the server, write a test suite from its real schemas, and run it — code decides pass/fail on the responses (schema conformance, error codes, pagination, result caps) while a scoring model grades plain-English tasks driven through the tools.
Get a pass/fail report per tool with the evidence behind each verdict — and replay the same suite after every schema change. Destructive tools are excluded from the run.
Run evalsOpen MCP Agent Studio with the connection pre-filled. Add your token, pick any of 60+ models, and start chatting — no install required.
Open Agent StudioCommon questions about connecting, scoping and using it safely.
Cashfree’s hosted MCP server at mcp.cashfree.com. It exposes orders, payments, refunds, settlements, payment links and beneficiary data as MCP tools for an AI assistant.
Only if the credential permits it. This is the one capability to think hard about: a refund is real money leaving, it is not reversible, and an agent acting on a misread instruction is an expensive mistake. Read-only credentials for anything investigative.
Use sandbox credentials. Cashfree keeps the environments fully separate, so a sandbox key simply cannot reach live transactions — which makes it the right default while you are working out what to let an agent do.
That is the strongest fit. "What happened to this payment" requires joining order, payment and settlement records, and an agent does that in one turn instead of three dashboard lookups by a support rep.
Similar shape, different rails. Cashfree is built around Indian payment methods — UPI, netbanking, local mandates — so its failure reasons and settlement model reflect those. Stripe is the better fit for card-led international flows.