Roles, Responsibilities & Adoption

Roles in an MCP ecosystem

Role	Responsibilities
Server developer	Implements tools/resources/prompts, input validation, clear descriptions, error handling
Host/client developer	Integrates MCP clients, builds consent UX, translates tool definitions for the model
Platform / DevOps engineer	Deploys and scales servers, manages transports, credentials, and uptime
Security engineer	Reviews servers, defines permission policies, monitors for abuse
Governance / compliance	Approves server inventory, audit requirements, data-handling policies
End user	Grants/denies consent, selects prompts, supervises agent actions

Tip

A recurring exam theme: the host enforces consent and security policy, while the server owns correct tool behavior and input validation. Know which side of the trust boundary each responsibility sits on.

Adoption path for organizations

1. Evaluate — identify high-value integrations (internal APIs, knowledge bases, dev tooling)

2. Pilot — start with read-only, low-risk servers in a controlled host

3. Govern — establish a vetted server inventory, consent policies, and audit logging

4. Scale — roll out to more teams; add write-capable tools with human-in-the-loop controls

5. Operate — monitor usage, review permissions, keep servers updated

flowchart LR
E[Evaluate] --> P[Pilot<br/>read-only servers] --> G[Govern<br/>vetted inventory] --> S[Scale<br/>write tools + HITL] --> O[Operate<br/>monitor and review]

Organizational adoption flow

Adoption signals to know

• Broad client support: the official MCP site highlights AI assistants such as Claude and ChatGPT, plus developer tools like Visual Studio Code and Cursor, as MCP hosts — “build once and integrate everywhere”
• Official SDKs in 10 languages (TypeScript, Python, C#, Go, Java, Rust, Swift, Ruby, PHP, Kotlin), maintained in the modelcontextprotocol GitHub organization
• A large ecosystem of official integrations and community servers exists for common systems (GitHub, databases, cloud providers), listed in the official servers repository
• An official MCP Registry (still in preview) provides standardized metadata for publicly accessible servers, backed by Anthropic, GitHub, PulseMCP, and Microsoft
• Open governance: Anthropic donated MCP to the Linux Foundation’s Agentic AI Foundation (AAIF) in December 2025, reinforcing vendor neutrality; spec changes are proposed via SEPs (Specification Enhancement Proposals)

Key exam points

• Consent UX is a host developer responsibility; input validation is a server developer responsibility
• Recommended adoption starts with low-risk, read-only use cases
• Governance requires a maintained inventory of approved servers
• MCP is governed openly under the Linux Foundation’s Agentic AI Foundation, not by a single vendor

Check your knowledge

📝 Check your knowledge

1. Who is responsible for validating tool inputs?

AThe host developerBThe server developerCThe end userDThe registry operator

2. Which role builds the consent UX that lets users approve or deny tool invocations?

AServer developerBPlatform engineerCHost/client developerDGovernance team

3. What kind of server is recommended for an initial organizational pilot?

AA write-capable automation server for maximum valueBA read-only, low-risk server in a controlled hostCA public community server with broad permissionsDA server that bypasses consent prompts to reduce friction

4. Which organization governs MCP today?

AIt is solely controlled by AnthropicBThe Agentic AI Foundation (AAIF), part of the Linux FoundationCThe OpenAI Developer ConsortiumDThe W3C

5. What must a governance/compliance function maintain for safe MCP adoption?

AA single shared API key for all serversBAn inventory of vetted, approved servers with audit requirementsCA copy of every model promptDDirect write access to all production servers