Complete reference for MCP tool integration across specialist and role agents including server inventory, tool recommendations, and security protocols
Status: Production Guide Last Updated: 2025-10-08 Purpose: Complete reference for MCP tool integration across specialist and role agents
This guide documents the complete MCP (Model Context Protocol) integration strategy for DA Agent Hub, ensuring all agents have comprehensive knowledge of their available tools and can make informed recommendations.
Specialists Recommend → Main Claude Executes → Specialists Analyze
Complete Tool Knowledge
Tool Selection Frameworks
| Server | Package | Tools | Primary Users | Status |
|---|---|---|---|---|
| dbt-mcp | dbt-mcp (uvx) | 40+ tools, 7 categories | dbt-expert, analytics-engineer-role | ✅ Connected |
| snowflake-mcp | snowflake-labs-mcp (uvx) | 26+ tools, 4 categories | snowflake-expert, dbt-expert, analytics-engineer-role | ✅ Connected |
| aws-api | awslabs.aws-api-mcp-server (uvx) | 3 core tools | aws-expert, frontend-developer-role, data-engineer-role | ✅ Connected |
| aws-docs | awslabs.aws-documentation-mcp-server (uvx) | 3 documentation tools | aws-expert | ✅ Connected |
| github | @modelcontextprotocol/server-github (npx) | 28 tools, 4 categories | github-sleuth-expert, documentation-expert, all role agents | ✅ Connected |
| slack | @modelcontextprotocol/server-slack (npx) | 8 tools | project-manager-role, business-analyst-role, qa-engineer-role | ✅ Connected |
| filesystem | @modelcontextprotocol/server-filesystem (npx) | 13 tools | github-sleuth-expert, documentation-expert, qa-engineer-role | ✅ Connected |
| sequential-thinking | @modelcontextprotocol/server-sequential-thinking (npx) | 1 cognitive tool | data-architect-role, qa-engineer-role, business-analyst-role | ✅ Connected |
MCP Access: dbt-mcp, snowflake-mcp, github-mcp, sequential-thinking-mcp Tool Count: 40+ dbt tools, 26+ Snowflake tools
Primary Tools:
Key Patterns:
Confidence: HIGH (0.92-0.95) for Discovery/Semantic/Admin tools
MCP Access: snowflake-mcp, dbt-mcp, sequential-thinking-mcp Tool Count: 26+ Snowflake tools
Primary Tools:
Key Patterns:
Confidence: HIGH (0.90-0.95) for core operations, MEDIUM (0.70-0.75) for Cortex
MCP Access: aws-api, aws-docs, sequential-thinking-mcp Tool Count: 6 tools across 2 MCP servers
Primary Tools:
call_aws (execute CLI), suggest_aws_commands (discover), get_execution_plan (experimental)read_documentation, search_documentation, recommend (related content)Key Patterns:
Critical Insight: aws-docs MCP provides current documentation (not just training data from January 2025) - essential for service limits, new features, best practices, security, API parameters
Confidence: HIGH (0.92) for aws-api, HIGH (0.92-0.95) for aws-docs
MCP Access: github-mcp, filesystem-mcp, sequential-thinking-mcp Tool Count: 28 GitHub tools, 13 filesystem tools
Primary Tools:
Key Patterns:
Known Issues: get_file_contents missing SHA (use push_files for updates)
Confidence: HIGH (0.88-0.95) for all operations except file updates (MEDIUM 0.65-0.70)
MCP Access: filesystem-mcp, github-mcp, dbt-mcp Tool Count: 13 filesystem tools, 28 GitHub tools
Primary Tools:
Key Patterns:
Confidence: HIGH (0.90-0.95) for filesystem, HIGH (0.85-0.88) for GitHub docs
When specialist agents provide MCP tool recommendations to main Claude:
### RECOMMENDED MCP TOOL EXECUTION
**Tool**: mcp__<server>__<tool_name>
**Parameters**:
- parameter1: "value1"
- parameter2: value2
- parameter3: [list, of, values]
**Expected Result**: Description of what this should return
**Success Criteria**: How to validate the operation succeeded
**Fallback**: Alternative approach if MCP unavailable
**Confidence**: HIGH (0.95) - Production-validated pattern
dbt-expert recommendation:
### RECOMMENDED MCP TOOL EXECUTION
**Tool**: mcp__dbt-mcp__get_model_details
**Parameters**:
- model_name: "fct_orders"
**Expected Result**: Compiled SQL, column definitions, dependencies
**Success Criteria**: Returns model metadata with all fields populated
**Fallback**: Read compiled SQL from target/ directory if MCP unavailable
**Confidence**: HIGH (0.95) - Core discovery operation
snowflake-expert recommendation:
### RECOMMENDED MCP TOOL EXECUTION
**Tool**: mcp__snowflake-mcp__run_snowflake_query
**Parameters**:
- statement: "SELECT * FROM SNOWFLAKE.ACCOUNT_USAGE.WAREHOUSE_METERING_HISTORY WHERE START_TIME >= DATEADD(day, -7, CURRENT_TIMESTAMP()) ORDER BY CREDITS_USED DESC LIMIT 10"
**Expected Result**: Top 10 warehouses by credit usage (last 7 days)
**Success Criteria**: Query returns warehouse names with credit totals
**Fallback**: Direct snowflake-connector-python if MCP unavailable
**Confidence**: HIGH (0.90) - Standard cost analysis query
aws-expert recommendation:
### RECOMMENDED MCP TOOL EXECUTION
**Tool**: mcp__aws-api__call_aws
**Parameters**:
- cli_command: "aws ecs describe-services --cluster my-cluster --services my-service --region us-west-2"
**Expected Result**: ECS service configuration and status
**Success Criteria**: Returns service ARN, status, task definition details
**Fallback**: AWS Console if MCP unavailable
**Confidence**: HIGH (0.92) - Standard infrastructure query
Increases Confidence:
Decreases Confidence:
| Server | Auth Method | Credentials | Security Controls |
|---|---|---|---|
| dbt-mcp | Service Token or PAT | DBT_TOKEN env var | SQL execution DISABLED by default |
| snowflake-mcp | OAuth/Password/Key Pair | SNOWFLAKE_PASSWORD env var | Read-only by default (SELECT, DESCRIBE, USE) |
| aws-api | AWS credentials | IAM/SSO/named profiles | READ_OPERATIONS_ONLY=true |
| aws-docs | None (public docs) | N/A | No auth required |
| github | Personal Access Token | GITHUB_PERSONAL_ACCESS_TOKEN | Scopes: repo, read:org, read:project |
| slack | Bot Token | SLACK_BOT_TOKEN | Scopes: channels:read, chat:write, users:read |
| filesystem | Allowed directories | N/A | Whitelist: /Users/TehFiestyGoat/da-agent-hub |
| sequential-thinking | None (cognitive tool) | N/A | No auth required |
dbt-mcp:
DISABLE_SQL=true unless explicitly neededsnowflake-mcp:
aws-api:
READ_OPERATIONS_ONLY=true restricts to read-onlygithub:
filesystem:
Use Case: Optimize slow dbt model
Workflow:
Confidence: HIGH (0.92) - Production-validated pattern
Use Case: Deploy new AWS service
Workflow:
Confidence: HIGH (0.90) - Documentation currency critical
Use Case: Analyze recurring dbt error across repositories
Workflow:
Confidence: HIGH (0.88) - Cross-tool coordination
Use Case: Assess and improve documentation across projects
Workflow:
Confidence: HIGH (0.90) - Documentation standards enforcement
| Role Agent | Primary MCP Access | Delegation Pattern |
|---|---|---|
| analytics-engineer-role | dbt-mcp, snowflake-mcp | Direct for simple queries, delegate complex to dbt-expert |
| data-engineer-role | github-mcp, filesystem-mcp | Direct for pipelines, delegate AWS to aws-expert |
| data-architect-role | sequential-thinking-mcp | Delegate tool-specific work to specialists |
| qa-engineer-role | filesystem-mcp, sequential-thinking-mcp | Direct for testing, delegate domain work to specialists |
| business-analyst-role | slack-mcp, sequential-thinking-mcp | Direct for communication, delegate technical to specialists |
| project-manager-role | slack-mcp, github-mcp | Direct for coordination, delegate technical to specialists |
| frontend-developer-role | github-mcp | Delegate AWS deployment to aws-expert |
| bi-developer-role | dbt-mcp (metrics), snowflake-mcp | Delegate complex analysis to dbt-expert, snowflake-expert |
When role agents should use MCP tools directly:
When role agents should delegate to specialists:
get_file_contents Missing SHAServer: github-mcp
Impact: Cannot reliably update files with create_or_update_file
Workaround: Use push_files for batch updates OR list_commits to get SHA
Confidence: MEDIUM (0.60) - Functional workaround
Tracking: https://github.com/github/github-mcp-server/issues/595
Server: dbt-mcp
Impact: SQL execution tools can MODIFY data
Mitigation: DISABLED by default via DISABLE_SQL=true, requires PAT
Confidence: MEDIUM (0.65-0.70) when enabled - requires validation
Recommendation: Keep disabled unless explicitly needed
Server: snowflake-mcp Impact: Cortex AI tools require explicit setup in Snowflake Mitigation: Document when Cortex features are available Confidence: MEDIUM (0.70-0.75) - Requires Cortex service configuration
Server: aws-api Impact: Cannot modify infrastructure via MCP (read-only mode) Mitigation: Design pattern - specialists recommend, human implements Confidence: Appropriate restriction - aws-expert provides guidance, not execution
tableau-mcp (if available)
prefect-mcp (custom integration)
orchestra-mcp (custom integration)
dbt Cloud Admin API (enhanced dbt-mcp)
Confluence MCP (knowledge management)
.mcp.jsonclaude mcp list
# Test dbt-mcp
mcp__dbt-mcp__list_metrics
# Test snowflake-mcp
mcp__snowflake-mcp__list_objects object_type="table" database_name="ANALYTICS_DW"
# Test aws-api
mcp__aws-api__call_aws cli_command="aws sts get-caller-identity"
# Test github
mcp__github__search_repositories query="org:your-org" perPage=5
knowledge/mcp-servers/.claude/agents/specialists/.claude/agents/roles/.claude/rules/ and .claude/skills/reference-knowledge/.claude/skills/reference-knowledge/agent-mcp-integration-guide/SKILL.mdGuide created: 2025-10-08 Author: Claude Code - MCP Integration Research Purpose: Complete MCP integration reference for DA Agent Hub agents