agents-mcp
Set up Model Context Protocol servers to give Claude and Codex access to databases, APIs, GitHub, and custom tools with fine-grained permission controls. Choose from official servers for PostgreSQL, SQLite, and popular platforms, or build custom TypeScript servers for internal integrations.
agents-mcp configures MCP servers that extend Claude and Codex with secure database, API, and custom tool access.
AI-generated summary based on this skill's SKILL.md
Install
vasilyu1983/AI-Agents-public/agents-mcp · repository language: Python
git clone https://github.com/vasilyu1983/AI-Agents-public
cp -r AI-Agents-public/frameworks/shared-skills/skills/agents-mcp ~/.claude/skills/agents-mcpnpx skillfed install vasilyu1983/AI-Agents-public/agents-mcpFrequently asked questions
AI-generated answers based on this skill's SKILL.md and metadata
How do I set up MCP servers for Claude?
agents-mcp provides a complete setup guide for deploying Model Context Protocol servers that extend Claude's capabilities. Start by installing the MCP SDK, configuring your server in Claude's settings file, and choosing your transport layer (stdio or HTTP). Official servers for PostgreSQL, SQLite, GitHub, and other platforms are available; select the ones matching your integration needs, then authenticate and test connectivity before production use.
How can I connect Claude to PostgreSQL via MCP?
agents-mcp includes PostgreSQL MCP server configuration for secure database access. Install the PostgreSQL server package, provide connection credentials (host, port, database, user), and configure scoped permissions to limit which tables and operations Claude can access. Test the connection using MCP health checks, then enable in your Claude client settings to query and manipulate data safely.
How do I build a custom MCP server in TypeScript?
agents-mcp supports building custom MCP servers for proprietary tools and internal APIs using TypeScript. Use the MCP SDK to define your server's tools, resources, and prompts; implement request handlers for each capability; choose stdio or HTTP transport; and deploy alongside your agent. The quick-start template includes authentication, error handling, and logging patterns for production readiness.
What permission controls does agents-mcp provide for tool access?
agents-mcp enables fine-grained permission management through allow/deny rules and scoped filesystem access. Define which tools Claude can invoke, restrict database queries to specific tables, limit file operations to designated directories, and audit all tool invocations. Security guardrails prevent unauthorized data exposure and enforce least-privilege access for both built-in and custom MCP servers.
How do I troubleshoot MCP connection and authentication failures?
agents-mcp provides diagnostics for common MCP issues: verify transport setup (stdio vs HTTP), check authentication credentials and token expiration, review server logs for initialization errors, and run health checks to confirm connectivity. Connection failures often stem from misconfigured credentials, network isolation, or missing server dependencies—agents-mcp documentation includes resolution steps for each scenario.
When should I use MCP instead of direct API calls in my agent?
agents-mcp's Model Context Protocol offers advantages over direct API integration: centralized permission management, standardized tool discovery, built-in authentication handling, and audit logging. Use MCP for production agents needing security controls, multi-service orchestration, or compliance requirements; direct calls suit simple prototypes or single-service integrations where overhead isn't justified.
SKILL.md
rendered from the published skill — quoted content, verbatim
MCP (Model Context Protocol) — Advisor & Reference
Specification: https://modelcontextprotocol.io/specification/2025-11-25 (November 2025)
When to Use MCP (Decision Tree)
| Scenario | Use MCP? | Why |
|---|---|---|
| Query PostgreSQL/MySQL/SQLite | Yes | Official servers exist, read-only by default |
| Access filesystem outside workspace | Yes | Scoped allowlists, audit trail |
| GitHub/Linear/Slack/Notion integration | Yes | Vendor MCP servers available |
| One-off HTTP API call | No | Use WebFetch or Bash curl |
| Internal API with auth | Maybe | Build custom MCP server if repeated, otherwise direct call |
| Need write access to production DB | Caution | Prefer read-only; if writes needed, scope tightly |
Rule of thumb: Use MCP when (1) an official/community server exists, (2) you need audit/permission control, or (3) you'll reuse the integration across sessions.
Quick Start (Local stdio via npx)
1) Create or edit
(truncated - see the full file via the links below)
Read as markdown · JSON record · Browse the source repository
File tree — 10 files
frameworks/shared-skills/skills/agents-mcp/SKILL.md
frameworks/shared-skills/skills/agents-mcp/assets/api/template-mcp-api.md
frameworks/shared-skills/skills/agents-mcp/assets/database/template-mcp-database.md
frameworks/shared-skills/skills/agents-mcp/assets/deployment/template-mcp-docker.md
frameworks/shared-skills/skills/agents-mcp/assets/filesystem/template-mcp-filesystem.md
frameworks/shared-skills/skills/agents-mcp/data/sources.json
frameworks/shared-skills/skills/agents-mcp/references/mcp-custom.md
frameworks/shared-skills/skills/agents-mcp/references/mcp-patterns.md
frameworks/shared-skills/skills/agents-mcp/references/mcp-security.md
frameworks/shared-skills/skills/agents-mcp/references/mcp-servers.md