MCP Integrations

Model Context Protocol (MCP) servers extend Claude Code's capabilities by giving access to external tools and data sources. Specialist Agent works fully without any MCP - all agents operate using local files and built-in tools. MCPs are optional enhancements.

What MCPs Add

MCP	What it does	Who benefits
Context7	Fetches up-to-date library documentation	You (the developer) when asking about APIs
Azion	Generates edge configs, deploys static sites	`@starter` and `@cloud` agents (Edge Mode)

Context7 - Library Documentation

What it does: Fetches up-to-date documentation and code examples for any programming library (Vue 3, React, Pinia, TanStack Query, etc.).

Who benefits: Primarily you, the developer. When you ask Claude Code about a library API, Context7 provides current docs instead of relying on training data. Agents don't query Context7 automatically - they follow your project's ARCHITECTURE.md and local conventions.

Configuration:

json

{
  "mcpServers": {
    "context7": {
      "type": "http",
      "url": "https://mcp.context7.com/mcp"
    }
  }
}

Already Included

Context7 is pre-configured in Specialist Agent's .mcp.json. No setup needed.

Azion - Edge Deployment & Configuration

What it does: Connects Claude Code to the Azion Edge Platform, giving agents access to Azion's documentation, code samples, CLI commands, API specs, Terraform configs, and a static site deployment tool.

Which agents use it:

@starter - After scaffolding, can generate Azion edge configs and deploy static sites
@cloud - Edge Mode uses Azion MCP tools to generate rules engine configs, Terraform resources, and observability queries

Available MCP Tools

The Azion MCP exposes 9 tools - 7 for search/generation and 1 for deployment:

Tool	Category	What it does
`search_azion_docs_and_site`	Search	Full-text search across Azion documentation
`search_azion_code_samples`	Search	Code samples for edge functions and frameworks
`search_azion_cli_commands`	Search	CLI syntax and usage for any operation
`search_azion_api_v3_commands`	Search	API v3 endpoints, payloads, examples
`search_azion_api_v4_commands`	Search	API v4 endpoints (latest)
`search_azion_terraform`	Search	Terraform provider resources and HCL examples
`create_rules_engine`	Generator	Generates Rules Engine configs (cache, routing, redirects)
`create_graphql_query`	Generator	Builds GraphQL queries for analytics and observability
`deploy_azion_static_site`	Deploy	Deploys a static site to Azion Edge

How it works in practice

For static sites (SSG output from Vite, Nuxt, Next.js, SvelteKit), the agent can deploy directly via deploy_azion_static_site.

For dynamic apps (edge functions, SSR), the agent generates the correct azion.config.js, CLI commands, and infrastructure configs - you run azion deploy yourself.

Configuration:

json

{
  "mcpServers": {
    "azion": {
      "type": "http",
      "url": "https://mcp.azion.com",
      "headers": {
        "Authorization": "Bearer <your-azion-personal-token>"
      }
    }
  }
}

Or via Claude Code CLI:

bash

claude mcp add --transport http azion https://mcp.azion.com \
  --header "Authorization: Bearer $AZION_PERSONAL_TOKEN"

Authentication Required

You need an Azion Personal Token. Create one in the Azion Console under Account Menu > Personal Tokens. Store it as an environment variable - never commit tokens to your repository.

Full Configuration Example

Here's a complete .mcp.json with all recommended servers:

json

{
  "mcpServers": {
    "context7": {
      "type": "http",
      "url": "https://mcp.context7.com/mcp"
    },
    "azion": {
      "type": "http",
      "url": "https://mcp.azion.com",
      "headers": {
        "Authorization": "Bearer <your-azion-personal-token>"
      }
    }
  }
}

Place this file at your project root as .mcp.json. Claude Code loads it automatically.

Agent + MCP Interaction Examples

@starter deploying to Azion Edge

bash

"Use @starter to create a Vue 3 app and deploy it to Azion Edge"

After scaffolding, the starter asks where you want to deploy. If you choose Azion and the MCP is available, it queries search_azion_code_samples for the correct Vite bundler config, generates azion.config.js, and deploys the static build via deploy_azion_static_site.

@cloud configuring edge infrastructure

bash

"Use @cloud to set up edge caching and routing rules for my API on Azion"

In Edge Mode, the cloud agent uses create_rules_engine to generate cache rules and routing behaviors, search_azion_terraform for IaC resources, and create_graphql_query for observability dashboards.

Using Context7 for library lookups

bash

"How do I configure staleTime in TanStack Vue Query v5?"

With Context7 available, Claude fetches the latest TanStack Query docs instead of relying on training data - ensuring you get current API signatures and examples.

MCP Discovery Hook

Specialist Agent includes a MCP Discovery native hook that runs on session start. It automatically:

Reads your .mcp.json configuration
Maps configured MCPs to the agents that benefit from them
Injects availability context so agents know what external tools they can use

This is enabled by default during npx specialist-agent init. No manual configuration needed.

Agent-MCP Capability Matrix

MCP Server	Enhanced Agents	What It Enables
Context7	`@builder`, `@doctor`, `@migrator`, `@deps`	Up-to-date library documentation
Azion	`@starter`, `@cloud`	Edge deployment, Rules Engine, Terraform
Playwright	`@tester`, `@doctor`, `@debugger`, `@perf`	Browser screenshots, e2e tests, DOM inspection
GitHub	`@sentry-triage`, `@reviewer`, `@ripple`, `@deps`, `@devops`	Create issues/PRs, read CI status, manage releases
Sentry	`@sentry-triage`, `@doctor`, `@debugger`	Fetch errors directly, no CLI required
Linear	`@planner`, `@analyst`, `@product`, `@orchestrator`	Read/create issues, link tasks to code
Jira	`@planner`, `@analyst`, `@product`, `@orchestrator`	Read/create issues, link tasks to code
Supabase	`@data`, `@builder`, `@security`	Read schemas, manage migrations, configure auth
Slack	`@orchestrator`, `@support`	Notifications on reviews, deploys, errors

Example Session Output

When MCPs are detected, the hook injects context like:

[MCP Discovery] MCPs available: context7, playwright | Enhanced agents: @builder(+context7) @tester(+playwright) @doctor(+context7,playwright)

Complementary Tools

These are not MCP servers bundled with Specialist Agent, but external tools that pair well with it.

Browser Automation

Tool	What It Does	Best For
Playwright MCP	Automated browser testing, screenshots, page interaction	`@tester` e2e tests, `@doctor` visual debugging
Chrome DevTools MCP	Console logs, network tab, DOM inspection	`@debugger` live debugging, `@perf` runtime analysis

These MCPs let agents see what's happening in the browser — console errors, network failures, rendering issues — without you having to copy-paste logs.

Playwright MCP setup:

json

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@anthropic/mcp-playwright"]
    }
  }
}

Voice Prompting

Voice input tools let you describe tasks verbally instead of typing long prompts:

Tool	Platform	How It Helps
Wispr Flow	macOS	Dictate complex prompts naturally — useful with `@analyst` interviews
SuperWhisper	macOS	Offline voice-to-text with custom vocabulary
Claude Voice Mode	Built-in	`/voice` command in Claude Code

Voice is especially productive for brainstorming sessions with /brainstorm and requirement interviews with @analyst.

Terminal Multiplexing

For parallel agent workflows, a terminal multiplexer lets you run multiple Claude Code sessions side by side:

Tool	What It Does
tmux	Split terminal into panes, each running a Claude Code session
iTerm2	macOS terminal with native split panes and session management
Ghostty	Fast GPU-accelerated terminal with splits
Windows Terminal	Native Windows terminal with tab/pane support

This pairs well with @orchestrator workflows — run the orchestrator in one pane and watch subagent activity in others.

MCP Integrations ​

What MCPs Add ​

Context7 - Library Documentation ​

Azion - Edge Deployment & Configuration ​

Available MCP Tools ​

Full Configuration Example ​

Agent + MCP Interaction Examples ​

@starter deploying to Azion Edge ​

@cloud configuring edge infrastructure ​

Using Context7 for library lookups ​

MCP Discovery Hook ​

Agent-MCP Capability Matrix ​

Example Session Output ​

Complementary Tools ​

Browser Automation ​

Voice Prompting ​

Terminal Multiplexing ​

MCP Integrations

What MCPs Add

Context7 - Library Documentation

Azion - Edge Deployment & Configuration

Available MCP Tools

Full Configuration Example

Agent + MCP Interaction Examples

@starter deploying to Azion Edge

@cloud configuring edge infrastructure

Using Context7 for library lookups

MCP Discovery Hook

Agent-MCP Capability Matrix

Example Session Output

Complementary Tools

Browser Automation

Voice Prompting

Terminal Multiplexing