Contrast MCP Server

The hosted server is read-only and is available now for Contrast SaaS.

Add the server to Claude Code by pointing it at your Contrast host followed by/mcp.

claude mcp add --transport http contrast-hosted-mcp https://app.contrastsecurity.com/mcp

Any MCP client that supports Streamable HTTP transport and OAuth 2.0 with PKCE can connect.

Your client discovers the OAuth configuration automatically through theWWW-Authenticateresponse header, which points to the standard/.well-known/oauth-protected-resourcemetadata document. Clients that support Dynamic Client Registration can register at/oauth2/connect/registeron the Contrast origin.

Support for more clients is in progress as their OAuth handling matures. If your client fails during OAuth registration before the login page appears, that is usually a client compatibility issue rather than a problem with your account.

The hosted server changes how access works without changing what you are allowed to see.

The shared warning above still applies. Tool results become part of your AI conversation, so follow your organization's policy on what security data can be sent to your chosen AI client and model.

The hosted server provides read-only tools across the domains below. Your agent calls them automatically based on your questions.

CVEs, Issues, Incidents, and Observations

These tools require the Contrast unified data platform (NorthStar) to be enabled for your organization.

The Local MCP Server is the open-source server in this repository. Your MCP client launches it as a local process over stdio, it authenticates with Contrast API and service keys, and it connects to your own Contrast instance, including on-premises and EOP. Use it when you cannot use the hosted server, or when you need raw SARIF scan output.

The Local MCP Server provides 14 tools for security analysis and vulnerability management.

VS Code (GitHub Copilot) - One-Click Install

Add this to yourmcp.jsonconfiguration file and replace the placeholder values with your Contrast credentials:

These prompts work with either server, except where a section says otherwise.

- Please list vulnerabilities for Application Y.
- Give me details about vulnerability X in Application Y.
- Review vulnerability X and fix it.

- Which libraries in Application X have high or critical vulnerabilities and are actively used?
- Update library X, which has a critical vulnerability, to the safe version.
- Which libraries in Application X are not being used?

- Please give me the applications tagged with "backend."

- Please give me the applications with metadata "dev-team" and "backend-team."

Retrieve Vulnerabilities by Session Metadata

- Give me the session metadata for Application X.
- Give me the vulnerabilities in the latest session for Application X.
- Give me the vulnerabilities for session metadata "Branch Name" "feature/some-new-fix" for Application X.
- Give me the route coverage for the latest session for Application X.
- Give me the route coverage for session metadata "Branch Name" "feature/some-new-fix" for Application X.

- Please give me a breakdown of applications and servers vulnerable to CVE-xxxx-xxxx.
- Please list the libraries for the application named xxx and tell me what version of commons-collections is being used.
- Which vulnerabilities in Application X are being blocked by a Protect or ADR rule?
- Which production servers do not have Protect enabled?
- Show me servers whose agents are out of date.
- Show me attack events from the last 7 days and tell me which were exploited.

Hosted server with the unified data platform (NorthStar)

These prompts require the hosted server and the Contrast unified data platform (NorthStar) to be enabled for your organization.

- Which CVEs pose the highest risk across my organization?
- Is my organization shielded from CVE-xxxx-xxxx?
- Which applications and libraries are affected by CVE-xxxx-xxxx?
- Show me open security issues for Application X.
- Give me the details of incident X and the issues linked to it.
- What observations provide evidence for issue X?

The Contrast MCP Server provides a bridge between your Contrast Data and the AI Agent/LLM of your choice. By using Contrast's MCP server you will be providing your Contrast Data to your AI Agent/LLM, it is your responsibility to ensure that the AI Agent/LLM you use complies with your data privacy policy. Depending on what questions you ask the following information will be provided to your AI Agent/LLM.

- Application Details
- Application Rule configuration
- Vulnerability Details
- Route Coverage data
- ADR/Protect Attack Event Details
- Server inventory and agent details (hostnames, paths, agent versions, environments, log levels, and tags)
- Issue, incident, observation, and CVE Shield data (hosted server with NorthStar)

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

Skill router and context picker for coding agents — hybrid retrieval + rerank picks the right skill; an ONNX prompt-injection gate scans both the request and every retrieved doc.

An intelligent security agent that analyzes AI-generated code for vulnerabilities and assists with remediation.

An MCP server to help AI assistants to answer questions and generate AccelByte Extend SDK code more effectively .

Instead of direct calling MCP tools, mcpcode server transforms MCP tool calls into TypeScript programs, enabling smarter, lower-latency orchestration by LLMs.

Supercharge your Agent with Semantic Code Intelligence and save 💰 in the process!

AI-to-AI code review platform — Claude, Codex, and Gemini cross-check each other via MCP, REST API, and CLI for consensus-based results.

A code sandbox for AI assistants to safely execute arbitrary code. Requires a 302AI API key for authentication.

Agent-native developer Q&A API with MCP + A2A endpoints for citations, job pickup, and answer submission.

Remote MCP server (Streamable HTTP) at https://mcp.agenticrail.nz/ — deterministic step-order enforcement for AI agents. evaluate_step returns ALLOW or DENY before a step runs; verify_receipt proves a sequence's Ed25519-signed, hash-chained receipt chain is intact. No auth required: omit the bearer token and calls run on the public demo key. That first clause matters — the form has no "remote/hosted" field, and putting the endpoint in the description is the convention on that list ("Fully REMOTE! Just use…"). The rest mirrors your own server card verbatim, so the listing and the card can't drift.