Cypress MCP server: 11 Cloud tools, capped hourly

A free connector with an unusually clear rate limit, tools that only work if you pay for the product behind them, and a documented privacy boundary.

Last verified 1 September 2026 · from Cypress's own Cloud MCP documentation

This page is one of 90 in a directory of monitoring and developer tools. Each page checks what an AI coding agent can get out of the tool through its MCP server, the connector that lets the agent query the tool directly.

What is the Cypress MCP server?

It is Cypress's server for the Model Context Protocol. Cypress Cloud records test runs, so a connected agent can read a CI failure without reproducing it locally.

The workflow the documentation describes runs in three steps. Find the project with cypress_get_projects, then the run with cypress_get_runs, then the failure with cypress_get_failed_tests.

That last tool returns the test name, the spec, the error and stack trace, and a Test Replay link. It does not return the run's stored artifacts, such as screenshots or videos.

Six of the eleven cover two separate products. Three read accessibility reports and three read UI Coverage, each with a summary tool, a views tool and a detail tool.

Cypress exposes runs, failures, flake, accessibility reports and UI Coverage from its hosted server. Those records connect test work to CI, while leaving production telemetry, agent traces and billing outside the server.

For an agent, Cypress answers what the browser tests did and where the interface is untested. It can read a failure and hand back a replay link. It cannot run a test, retry a build or say anything about production, so it works beside a pipeline connector rather than instead of one.

What can an AI agent do with Cypress?

Both report tools return only the top five views, the distinct screens Cypress recorded, ranked by failure count. Reaching the sixth needs the paging tool, and the report gives no sign that five was a cap.

What connecting costs before the first question

A context window is the amount of text a model can hold at once. Eleven tools ship on every Cypress plan at no cost, but they are not free of context: 1,211 to 1,284 tokens on Claude before a single test result is fetched.

Read this next to the rate limit rather than on its own. The connection is capped at 100 tool requests an hour, so the context is spent once per session and the cap decides how much use you get from it.

Encodings: GPT-5.6, GPT-5.5 o200k_base; GPT-4, GPT-3.5 cl100k_base; Claude Haiku 4.5 to Opus 5 count_tokens. Counted from the 11 tool entries in Cypress's own documentation. That is the only public inventory.

An unauthenticated tools/list against mcp.cypress.io/mcp returns 401. So no input schemas could be captured. These figures cover each tool's name and description only. They are a floor, not the figure.

The Claude columns are measured through the API's count_tokens endpoint, with an empty schema per tool. So they cover the same content as the GPT columns. The range there is two tokenizer generations. Haiku 4.5 and Opus 4.6 sit at the low end. Opus 5 and Sonnet 5 sit at the high end. The gap between them is about a tenth.

Cypress MCP server limits: where answers come back incomplete

Both limits below produce the same result: an empty answer that reads like good news.

How to configure Cypress MCP for agents

1. 1

   An administrator enables it once

   The integration is switched on from the organization's Integrations page before anyone can authenticate.

2. 2

   Each person authenticates separately

   OAuth sessions are scoped to the individual account and respect that person's role in each organization.

3. 3

   Plan for 100 requests an hour

   That is the documented cap on every plan, and an agent that pages through views can reach it in one investigation.

4. 4

   Know which tools your plan feeds

   Flake, accessibility and UI Coverage tools return data only where the organization subscribes to those products.

Paste this into your agent’s instructions

Two of the tools return only the top five views, and several return nothing without a subscription. Both look like an absence of problems.

When you answer using Cypress Cloud MCP tools, state:
- The project and run the answer covers.
- Whether a report was truncated. The accessibility and UI
  Coverage reports return only the top five views.
- Whether a tool returned nothing because the organization
  does not subscribe to the product behind it.
Never state that data does not exist. State that the run
above returned no results for that tool.
Write your answer in ASD-STE100 Simplified Technical
English. Use short sentences with one idea in each.

Do you need the MCP server at all?

Not always. Cypress documents a second route itself. The Cloud CLI, cy-cloud, returns JSON from a terminal and reaches Test Replay data directly, which the documentation recommends for scripting and deeper triage.

The connector earns its place in a conversation. The agent picks the tool from the prompt, which suits in-editor debugging rather than a scripted pipeline.

Cypress's own suggestion is that many teams use both, choosing by whether the assistant speaks the protocol or runs shell commands.

Connect the server for in-editor debugging. Use the Cloud CLI for scripting and terminal triage.

Cypress MCP server setup

The server is hosted, so there is nothing to install. An administrator enables the integration for the organization from the Integrations page.

Each person then authenticates. OAuth is the documented recommendation; a personal access token exists for clients that cannot open a browser.

claude mcp add --transport http cypress \
  https://mcp.cypress.io/mcp

{"mcpServers":{"cypress":{
  "url":"https://mcp.cypress.io/mcp"}}}

[mcp_servers.cypress]
url = "https://mcp.cypress.io/mcp"

Every other client

Each block below is the configuration for one client, with the file path and the key that client expects.

{"mcpServers":{"cypress":{"url":"https://mcp.cypress.io/mcp"}}}

{"mcpServers":{"cypress":{"url":"https://mcp.cypress.io/mcp"}}}

{"servers":{"cypress":{"url":"https://mcp.cypress.io/mcp"}}}

{"mcpServers":{"cypress":{"url":"https://mcp.cypress.io/mcp"}}}

{"mcpServers":{"cypress":{"command":"npx",
  "args":["-y","mcp-remote","https://mcp.cypress.io/mcp",
    "--header","Authorization: Bearer $CYPRESS_MCP_TOKEN"],
  "env":{"CYPRESS_MCP_TOKEN":"<your-token>"}}}}

Read-only access, permissions and security

• Cypress reports completed work

  Nothing in the documented list starts a run, changes a project or deletes a record.

• Access follows the person

  Both OAuth sessions and personal access tokens are scoped to the individual's role and permissions in each enabled organization.

• Cypress states what it receives

  Its documentation says it has no access to the assistant's context window or chat history, and that only the tool call's inputs reach it.

• Sessions are revocable

  Active OAuth sessions appear on the Cypress Cloud profile page and can be revoked there.

Troubleshooting

The client cannot connect

An administrator has to enable the Cloud MCP integration for the organization from its Integrations page before anyone can authenticate.

A tool returns nothing on a real run

Flake, accessibility and UI Coverage tools return plan-specific and product-specific data. Without the subscription behind them they have nothing to return.

Calls start failing during an investigation

The documented limit is 100 tool requests an hour on every plan. Paging through views consumes them quickly; Cypress directs requests for higher limits to its support address.

The client cannot complete a browser sign-in

Generate a personal access token from the Cypress Cloud profile page, under MCP personal access token. It is displayed once at creation.

Cypress MCP server: Reference

What engineers report

Cypress publishes limits and availability plainly, so the counts below come from its own documentation.

Tool inventory, limits and authentication read from Cypress's own Cloud MCP documentation on 24 August 2026. An unauthenticated tools/list against mcp.cypress.io/mcp returned 401 on the same date, which is why this page is sourced from documentation rather than from the wire.

Should you connect an agent to Cypress?

Yes. Best for OAuth from an interactive client, with a personal access token kept for agents that cannot open a browser.

Strongest fit

Cypress Cloud teams debugging CI failures from the editor rather than reproducing them locally.

Main advantage

Free on every plan, with a published hourly limit and a documented privacy boundary.

Main weakness

Seven of the eleven tools depend on a separate product or plan, and return empty rather than saying so.

Operational risk

None from the tools themselves. Every one reads, and access follows the person's own role.

Frequently asked questions