agentsSource-backedReview first Safety ✓ Privacy ✓
MCP Integration Architect Agent
Reusable agent role for MCP integration architecture decisions: choose local versus remote servers, pick documented transports, scope project versus user configuration, and minimize tool surface before rollout.
HarnessClaude Code
Review first — review before installing
Open the source and read safety notes before installing.
Safety notes
- MCP servers can expose write tools, network egress, and account-backed APIs.
- Project-scoped servers require explicit approval before use in shared repositories.
- Prefer read-only tools during pilot rollout; expand scope only after review.
Privacy notes
- MCP tool outputs may include repository paths, account metadata, or customer data.
- OAuth flows may expose tenant identifiers—redact before sharing architecture notes externally.
Prerequisites
- Description of the workflow needing tools or external data.
- Access to review .mcp.json or claude mcp list output in a staging environment.
- Security reviewer for project-scoped server approval and OAuth flows.
Schema details
- Install type
- copy
- Troubleshooting
- No
Source repository stats
- Scope
- Source repo
Full copyable content
You are an MCP Integration Architect Agent.
Help teams decide whether MCP is the right integration path and how to configure
servers using documented Claude Code MCP workflows.
Workflow:
1. Restate the task, data boundaries, and whether tools need local files or remote APIs.
2. Compare MCP versus simpler alternatives (slash commands, scripts, direct API clients).
3. If MCP fits, choose transport using official guidance:
- HTTP for remote cloud services (recommended)
- stdio for local command-based servers
- avoid deprecated SSE unless required by a legacy server
4. Pick configuration scope (project, user, or local) and document approval needs
for project-scoped .mcp.json entries.
5. Draft minimal tool exposure: prefer read-only tools for first rollout.
6. Plan verification with claude mcp list, claude mcp get, and /mcp auth checks.
7. Document rollback with claude mcp remove and reverting .mcp.json edits.
Output:
- Fit decision (MCP yes/no and why)
- Recommended transport and scope
- Tool minimization plan
- Verification checklist
- Rollback steps
- Privacy notes for credentials and tool outputsAbout this resource
Role
This agent supports architecture decisions before MCP rollout. It applies documented Claude Code MCP management workflows; it is a community agent prompt, not an Anthropic-built agent product.
When to activate
- Evaluating whether a new external integration should be an MCP server
- Choosing HTTP versus stdio transport for a proposed server
- Planning project-scoped
.mcp.jsonentries that need maintainer approval - Minimizing tool surface area before enabling servers for a team
Architecture checklist
- Task requires tool calls or structured external data (not just chat context).
- Transport matches server deployment (HTTP for remote, stdio for local commands).
- Scope is intentional: project, user, or local per team policy.
- Verification plan uses
claude mcp list,claude mcp get, and/mcp. - Rollback documented with
claude mcp removeand config revert steps. - OAuth or remote servers have a non-production pilot credential plan.
Source Verification Notes
Verified against Claude Code MCP documentation on 2026-06-16:
- HTTP transport is the recommended option for remote MCP servers.
- SSE transport is documented as deprecated in favor of HTTP where available.
- stdio servers run local commands;
--separates Claude flags from server args. claude mcp add --scopestores configuration at project, user, or local scope.- Project-scoped
.mcp.jsonservers can show pending approval until reviewed interactively. claude mcp list,claude mcp get, andclaude mcp removemanage server lifecycle.
Duplicate Check
Distinct from claude-mcp-skills-integration-agent (Skills-oriented integration),
mcp-authorization-boundary-review-agent (OAuth boundary review), and
mcp-remote-server-security-auditor-agent (security audit focus). This agent covers
fit and architecture before installation.
References
- Claude Code MCP documentation: https://code.claude.com/docs/en/mcp
- MCP getting started: https://modelcontextprotocol.io/docs/getting-started/intro
Source citations
Add this badge to your README
Show that MCP Integration Architect Agent is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.
[](https://heyclau.de/entry/agents/mcp-integration-architect-agent)How it compares
MCP Integration Architect Agent side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
| Field | MCP Integration Architect Agent Reusable agent role for MCP integration architecture decisions: choose local versus remote servers, pick documented transports, scope project versus user configuration, and minimize tool surface before rollout. Open dossier | Claude Agent SDK MCP Integration Capability Pack Skill Expert Claude Agent SDK MCP integration capability pack for designing, reviewing, and rolling out Agent SDK MCP integration with source-backed checklists, production rules, and privacy-safe output contracts. Open dossier | MCP Remote Server Trust Review Capability Pack Skill Expert MCP remote server trust review capability pack for auditing OAuth flows, transport security, tool permissions, data exfiltration risk, and vendor scope before connecting Claude Code to third-party MCP servers. Open dossier | Claude Code managed MCP governance Capability Pack Skill Expert Claude Code managed MCP governance capability pack for designing, reviewing, and rolling out managed MCP governance with source-backed checklists, production rules, and privacy-safe output contracts. Open dossier |
|---|---|---|---|---|
| Trust | ||||
| Install risk | Review first | Review first | Review first | Review first |
| Notes | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ |
| Category | agents | skills | skills | skills |
| Source | source-backed | source-backed | source-backed | source-backed |
| Author | kiannidev | kiannidev | kiannidev | kiannidev |
| Added | 2026-06-16 | 2026-06-14 | 2026-06-14 | 2026-06-14 |
| Platforms | Claude Code | Claude CodeCodexWindsurfGeminiCursorCLI | Claude CodeCodexWindsurfGeminiCursorCLI | Claude CodeCodexWindsurfGeminiCursorCLI |
| Source repo | — | — | — | — |
| Safety notes | ✓MCP servers can expose write tools, network egress, and account-backed APIs. Project-scoped servers require explicit approval before use in shared repositories. Prefer read-only tools during pilot rollout; expand scope only after review. | ✓This skill plans Agent SDK MCP integration; it must not execute destructive changes without explicit approval. Browser, computer-use, and remote surfaces can access sensitive UI state; scope tests carefully. MCP and SDK integrations may exfiltrate data if tool scopes are too broad. The public `anthropics/claude-code` repository ships documentation links to code.claude.com for settings, security, and integration surfaces. Scheduled or autonomous workflows compound risk; cap blast radius in staging first. | ✓Remote MCP servers run outside Anthropic control; Claude Code MCP integration does not guarantee vendor security or data isolation. OAuth tokens issued to an MCP server may grant persistent access to third-party accounts until revoked in the vendor admin console. Tools that read, write, delete, or execute on external systems can cause irreversible production changes when invoked by the model. SSE and streamable HTTP transports must use TLS; do not approve cleartext remote endpoints on untrusted networks. This skill recommends scoping and approval steps; it must not add MCP servers or approve OAuth consent without explicit user authorization. | ✓This skill plans managed MCP governance; it must not execute destructive changes without explicit approval. Browser, computer-use, and remote surfaces can access sensitive UI state; scope tests carefully. MCP and SDK integrations may exfiltrate data if tool scopes are too broad. The public `anthropics/claude-code` repository ships documentation links to code.claude.com for settings, security, and integration surfaces. Scheduled or autonomous workflows compound risk; cap blast radius in staging first. |
| Privacy notes | ✓MCP tool outputs may include repository paths, account metadata, or customer data. OAuth flows may expose tenant identifiers—redact before sharing architecture notes externally. | ✓Reviews may expose integration tokens, customer metadata, and internal URLs related to Agent SDK MCP integration. Telemetry and analytics configs can include account emails; redact before sharing externally. Keep troubleshooting logs in internal channels unless explicitly sanitized. Third-party vendors remain outside Anthropic retention policies; document separately. | ✓MCP tool results can contain customer names, ticket contents, database rows, repository secrets, and internal URLs that should not be pasted into public issues. OAuth consent screens and server logs may expose account emails, organization identifiers, and access tokens if shared without redaction. Remote server vendors may retain prompts, tool arguments, and responses under their own privacy policies outside Anthropic data handling. Public trust-review summaries should describe risk categories and mitigations, not full tool schemas or live OAuth tokens. | ✓Reviews may expose integration tokens, customer metadata, and internal URLs related to managed MCP governance. Telemetry and analytics configs can include account emails; redact before sharing externally. Keep troubleshooting logs in internal channels unless explicitly sanitized. Third-party vendors remain outside Anthropic retention policies; document separately. |
| Prerequisites |
|
|
|
|
| Install | — | — | — | — |
| Config | — | — | — | — |
| Citations | ||||
| Claim | Unclaimed | Unclaimed | Unclaimed | Unclaimed |
Signals
Loading live community signals…
More like this, weekly
A short, calm digest of reviewed Claude resources. Unsubscribe any time.