Skip to main content
This page lists the policies that apply to everyone who takes part in the MCP project: in meetings, on Discord, and in every repository under the modelcontextprotocol organization. Group Leads and Facilitators link it from meeting agendas.

Code of Conduct

The Code of Conduct governs every channel, repository and meeting. Send reports to mcp-coc@anthropic.com. The Community Moderators handle them under the Contributor Ladder and escalate to Core Maintainers when needed.

Antitrust

Participants often work for competing organizations. MCP is organized as Model Context Protocol a Series of LF Projects, LLC, so the LF Projects policies, including its antitrust policy, apply to all participants. The project’s own Antitrust Policy sets out how meetings and standard-setting discussions are conducted.

Licensing and intellectual property

Code and specification contributions are made under the Apache License 2.0 and other documentation under CC-BY-4.0, with no copyright assignment. The governing text is General Project Policies.

AI-assisted contributions

Any AI assistance used for a contribution must be disclosed in the pull request or issue, together with its extent. This includes AI-written review responses and comments; trivial spacing or typo fixes are the only exception. The full policy is AI_POLICY.md, which applies to every repository in the organization.

Reporting a vulnerability

Do not report security issues in public issues, discussions or pull requests.
Open a private advisory on the affected repository at github.com/modelcontextprotocol/<repo>/security/advisories/new. Private vulnerability reporting is enabled on the specification repository and on every official SDK and project repository. Before filing, read SECURITY.md: it says what to include, describes the trust model and the behaviors that are intentional rather than vulnerabilities (stdio command execution, denial of service by a stdio peer, server side effects, LLM-driven tool invocation), and lists what remains in scope. A private advisory is also the project’s emergency security reporting mechanism. Report actively exploited vulnerabilities and severe incidents through a private advisory on the affected repository. State at the start of the report that the vulnerability is being actively exploited or that an incident is under way. The maintainers who receive a report coordinate the fix and advisory across every affected SDK so that advisories are released together, and raise it with the specification maintainers when the root cause is in the specification (see SDK Vulnerability Disclosure). CVEs are assigned through GitHub’s CNA. Questions that are not themselves a report, including which repository to file against, go to the Security Interest Group in #security-ig.

EU Cyber Resilience Act (CRA) stewardship

The Model Context Protocol project publishes software, including the official SDKs, that is ultimately intended for commercial activities. The project is supported under the Linux Foundation CRA stewardship framework. See the Linux Foundation’s CRA guidance for open source projects for more on what the CRA asks of projects like this one. Maintainers who learn of an actively exploited vulnerability or a severe incident, such as a compromise of the release process, must notify a Core Maintainer or Lead Maintainer immediately in #security-disclosures on Discord, while they fix the problem and never instead of fixing it. That Core Maintainer or Lead Maintainer contacts the Linux Foundation CRA steward at steward@linuxfoundation.org and works with the steward from then on, so the project speaks to the steward with one voice. The CRA sets short deadlines for the steward’s reports, starting with an early warning within 24 hours and a fuller notification within 72 hours. The advisory, if there is one, records when the project first learned of the problem and when the steward was notified. A vulnerability with no evidence of exploitation is not a CRA escalation event. Other maintainers do not contact the steward directly, and reporters must use a private advisory.

Communication and records

Communication sets out which channel is for what, when a channel may be private, and where decisions are recorded.