Integration & MCP Server Catalog

Maintained by: Ivan (Infrastructure Engineer) Last updated: 2026-03-29 Decision: TFD-0014 (Component Taxonomy — Category 7)

Overview

This catalog documents all available integrations for digital talents — MCP servers, external APIs, and tool connections. When Pablo assembles a talent and the work order specifies a client tool (e.g., "LeanIX", "Confluence"), this catalog tells him what's available and how to configure it.

MCP Servers

Tier 1: Production-Ready

These are tested, documented, and have been used in at least one shipped talent.

Server Package Purpose Auth Used In
Atlassian (Jira) claude.ai managed Issue CRUD, search (JQL), transitions, comments, worklogs, linking OAuth (Atlassian cloud) Factory operations, request tracking
Atlassian (Confluence) claude.ai managed Page CRUD, search (CQL), comments, space browsing OAuth (Atlassian cloud) STM EA (publication target), factory KB
Draw.io drawio-mcp (npm) Diagram generation and editing None (local) STM EA Agent (9 diagram templates)
Context7 claude.ai managed Documentation lookup for libraries, frameworks, APIs None Factory R&D, skill development

Tier 2: Available, Partially Tested

Server Package Purpose Auth Notes
Figma claude.ai managed Design context, screenshots, Code Connect, FigJam OAuth (Figma) Used in factory (web design); not yet in a product
Telegram Plugin (telegram@claude-plugins-official) Message reply, reactions, attachments Bot token Factory communication; not client-facing

Tier 3: Evaluated, Not Yet Deployed

Server Package Purpose Status
Paperclip paperclipai/paperclip Multi-agent orchestration platform R&D evaluation (intake 2026-03-28)

Tier 4: Known but Not Evaluated

Potential MCP servers for future work orders:

Server Potential Use Case Client Demand
SharePoint Publication target for Microsoft-shop clients Anticipated
Notion Publication target, project management Low
ServiceNow ITSM integration, incident management Anticipated (enterprise clients)
GitHub Code repo integration, PR management Low (current clients don't need)
Slack Client communication channel Anticipated
Google Workspace Docs/Sheets integration Low

Storage & Persistence MCP Servers

Dedicated section for storage integrations. See also: Harness Configuration Reference Section 7 for when to use each.

Server Type Status Use Case Reference
Mem0 MCP (OpenMemory) Universal memory layer (vector + graph) Tier 3 — Evaluated, recommended Cross-session memory, model-switching persistence RD-0011 (score 3.4)
Context Mode MCP SQLite-backed context virtualization Tier 3 — Evaluated, caveats Structured queries, FTS5 search, 98% context reduction RD-0008 (score 3.0, single maintainer risk)
LangGraph Persistence Checkpoint store (PostgreSQL/Redis/DynamoDB) Tier 4 — Known, not tested Multi-agent state checkpoints, workflow recovery RD-0011
mcp-memory-service Open-source MCP memory Tier 4 — Known, not tested Community alternative to Mem0 RD-0011

Gap: No vector store MCP evaluated yet (Chroma, Pinecone, Weaviate). Needed for semantic RAG implementation.

Non-MCP Integrations

These integrations don't use MCP but are configured through content-in, skills, or CLAUDE.md.

Tool-Specific Data Models

Tool Integration Method Documentation Used In
SAP LeanIX (v4) Data model docs in content-in/, extraction skills Framework package: frameworks/enterprise-architecture/leanix/ STM EA Agent
Macroscope Methodology docs in content-in/ Framework package: frameworks/enterprise-architecture/macroscope/ STM EA Agent
TOGAF Framework guide in content-in/ Framework package: frameworks/enterprise-architecture/togaf/ EA product line

File Format Integrations

Format Direction Skills Notes
Draw.io XML Write Diagram generation skills Via Draw.io MCP or direct XML generation
CSV Read/Write Data extraction, catalog export LeanIX catalog schemas
Markdown Read/Write All skills Native format
JSON Read/Write Config, data transformation API responses, structured data
PDF Read Reference material ingestion Claude can read PDFs natively

Integration Selection Guide

During Assembly (Stage 4)

Work order specifies "Repository tool: {tool}"?
  YES → Check this catalog:
    Tool has Tier 1 MCP? → Configure .mcp.json + allowlist in settings.json
    Tool has data model docs? → Load into content-in/ + build integration skills
    Tool not in catalog? → Flag to Ivan for evaluation
  NO → Skip integration setup (standalone talent)

Work order specifies "Publication target: {platform}"?
  YES → Check this catalog:
    Platform has Tier 1 MCP? → Configure for write access
    Platform has Tier 2 MCP? → Test in dev; upgrade if stable
    Platform not in catalog? → Flag to Ivan; consider manual export workflow
  NO → Default to markdown output

MCP Configuration Template

Each talent that uses MCP ships with .mcp.json:

{
  "mcpServers": {
    "server-name": {
      "command": "npx",
      "args": ["-y", "server-package"],
      "env": {
        "API_KEY": "${CLIENT_API_KEY}"
      }
    }
  }
}

And corresponding permissions in settings.json:

{
  "permissions": {
    "allow": [
      "mcp__server-name__method1",
      "mcp__server-name__method2"
    ]
  }
}

Atlassian Integration Details

The most extensively used integration. Documenting full capability for reference.

Jira Capabilities

Capability MCP Method Factory Use
Browse projects getVisibleJiraProjects Project discovery
Search issues searchJiraIssuesUsingJql Request sync, status checks
Create issue createJiraIssue Request creation (with hook validation)
Edit issue editJiraIssue Status updates, field changes
Get issue details getJiraIssue Context loading
Transition issue transitionJiraIssue Workflow state changes
Add comment addCommentToJiraIssue Progress notes
Add worklog addWorklogToJiraIssue Time tracking
Link issues createIssueLink Cross-reference
Get field metadata getJiraIssueTypeMetaWithFields Dynamic form building

Confluence Capabilities

Capability MCP Method Factory Use
Browse spaces getConfluenceSpaces Space discovery
Get page getConfluencePage Content retrieval
Create page createConfluencePage Documentation publishing
Update page updateConfluencePage Content updates
Search (CQL) searchConfluenceUsingCql Knowledge search
Comments createConfluenceFooterComment Review feedback
Page tree getConfluencePageDescendants Navigation

Hooks for Atlassian

Two PreToolUse hooks enforce quality on Jira operations:

  1. Naming convention hook — Validates issue summary format (verb-first, max 70 chars, Epic prefix codes)
  2. Content standards hook — Requires Acceptance Criteria + Definition of Done for Stories/Bugs

Maturity Advancement Process

How an integration moves from Tier 4 → Tier 1:

Stage Gate Who
Tier 4 → Tier 3 Identified in a work order or R&D scan Riley or Pablo
Tier 3 → Tier 2 Ivan tests: install, auth, basic operations work Ivan
Tier 2 → Tier 1 Used in a shipped talent, documented, hooks written if needed Ivan + Pablo

Update Triggers

Trigger Action
New work order with unknown tool Ivan evaluates; adds to catalog at appropriate tier
MCP server update/new release Ivan tests; updates tier if needed
Client feedback on integration Fiona reports; Ivan adjusts configuration
New MCP server announced Riley logs R&D intake; Ivan evaluates when prioritized
Integration breaks in production Ivan investigates; updates Known Issues

References

  • Assembly guide Phase A (MCP config): production-lines/digital-talent/assembly.md
  • Deployment manifest example: production-lines/orders/WO-0001-stm-ea-agent/output/deployment-manifest.md
  • Factory settings: .claude/settings.json
  • LeanIX framework: production-lines/digital-talent/frameworks/enterprise-architecture/leanix/
  • Harness config reference: departments/it-architecture/agentic-pattern-designer-ada/harness-configuration-reference.md