🔍 MCP Inspector Report - 2026-02-09

[github-actions[bot]]

Systematic investigation of all MCP server configurations in .github/workflows/shared/mcp/.

Summary

• Servers Inspected: 21
• By Type: Container: 11, HTTP: 5, Local: 2, Stdio: 2, Service: 1
• Secrets Required: 10 servers require authentication
• Safe Jobs: 2 custom safe-jobs (Slack, Notion)

Inventory Table

Server	Type	Tools	Secrets	Status
arxiv	Container	3	No	✅
ast-grep	Container	All	No	✅
azure	Container	All (40+)	Yes	✅
brave-search	Container	All	Yes	✅
chroma	Container	14	No	✅
context7	Container	2	Yes	✅
datadog	HTTP	4	Yes	✅
deepwiki	HTTP	3	No	✅
drain3	Local	8	No	✅
fabric-rti	Stdio	13	Yes	✅
jupyter	Service	All	No	✅
markitdown	Container	All	No	✅
microsoftdocs	HTTP	All	No	✅
notion	Container	4	Yes	✅
semgrep	Container	All (6)	No	✅
sentry	Stdio	14	Yes	✅
server-memory	Container	4	No	✅
skillz	Container	Dynamic	No	✅
slack	Safe-Job	N/A	Yes	✅
svelte	HTTP	5	No	✅
tavily	HTTP	All	Yes	✅

View Detailed Server Information

arxiv

• File: shared/mcp/arxiv.md
• Type: Container (mcp/arxiv-mcp-server)
• Tools: search_arxiv, get_paper_details, get_paper_pdf
• Secrets: None
• Notes: Access to arXiv research papers

ast-grep

• File: shared/mcp/ast-grep.md
• Type: Container (mcp/ast-grep:latest)
• Tools: All tools allowed (structural code analysis)
• Secrets: None
• Notes: Tree-sitter based code search for multiple languages

azure

• File: shared/mcp/azure.md
• Type: Container (mcr.microsoft.com/azure-sdk/azure-mcp:latest)
• Tools: All tools (40+ Azure services)
• Secrets: AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET
• Notes: Read-only mode with --read-only flag, covers Key Vault, Storage, SQL, Service Bus, RBAC, Monitor, Virtual Networks, AI Foundry, etc.

brave-search

• File: shared/mcp/brave.md
• Type: Container (docker.io/mcp/brave-search)
• Tools: All tools (web and local search)
• Secrets: BRAVE_API_KEY
• Notes: Web search using Brave Search API

chroma

• File: shared/mcp/chroma.md
• Type: Container (mcp/chroma)
• Tools: 14 tools (collections: 8, documents: 5, embeddings: 1)
  • Collection Management: chroma_list_collections, chroma_create_collection, chroma_peek_collection, chroma_get_collection_info, chroma_get_collection_count, chroma_modify_collection, chroma_delete_collection, chroma_fork_collection
  • Document Operations: chroma_add_documents, chroma_query_documents, chroma_get_documents, chroma_update_documents, chroma_delete_documents
  • Additional: mcp_known_embedding_functions
• Secrets: None
• Notes: Requires cache-memory with chroma cache ID for persistent vector database storage

context7

• File: shared/mcp/context7.md
• Type: Container (mcp/context7)
• Tools: query-docs, resolve-library-id
• Secrets: CONTEXT7_API_KEY
• Notes: Up-to-date library documentation from Upstash

datadog

• File: shared/mcp/datadog.md
• Type: HTTP (`(mcp.datadoghq.com/redacted)
• Tools: search_datadog_dashboards, search_datadog_slos, search_datadog_metrics, get_datadog_metric
• Secrets: DD_API_KEY, DD_APPLICATION_KEY, DD_SITE (optional, defaults to datadoghq.com)
• Notes: Observability and monitoring platform integration

deepwiki

• File: shared/mcp/deepwiki.md
• Type: HTTP (`(mcp.deepwiki.com/redacted)
• Tools: read_wiki_structure, read_wiki_contents, ask_question
• Secrets: None (public service)
• Notes: GitHub repository documentation and AI-powered Q&A

drain3

• File: shared/mcp/drain3.md
• Type: Local HTTP (`(localhost/redacted)
• Tools: index_file, query_file, list_templates, list_clusters, cluster_stats, find_anomalies, compare_runs, search_pattern
• Secrets: None
• Notes: Log template mining and pattern extraction, requires Python 3.11+, FastMCP 2.0, custom setup steps, streaming JSONL responses, state persistence in workspace

fabric-rti

• File: shared/mcp/fabric-rti.md
• Type: Stdio (uvx microsoft-fabric-rti-mcp)
• Tools: 13 tools (Eventhouse: 10, Eventstreams: 3)
  • Eventhouse: kusto_known_services, kusto_query, kusto_list_databases, kusto_list_tables, kusto_get_entities_schema, kusto_get_table_schema, kusto_get_function_schema, kusto_sample_table_data, kusto_sample_function_data, kusto_get_shots
  • Eventstreams: list_eventstreams, get_eventstream, get_eventstream_definition
• Secrets: AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET
• Notes: Microsoft Fabric Real-Time Intelligence, read-only access, excludes destructive operations (kusto_command, kusto_ingest_inline_into_table)

jupyter

• File: shared/mcp/jupyter.md
• Type: Service (Docker Compose with 2 services: jupyter, jupyter-mcp)
• URL: `(jupytermcp/redacted)
• Tools: All tools (execute cells, create/list notebooks, get outputs)
• Secrets: None (uses github.run_id as token)
• Notes: Full Jupyter notebook integration with image output support

markitdown

• File: shared/mcp/markitdown.md
• Type: Container (mcp/markitdown)
• Tools: All tools (document conversion to markdown)
• Secrets: None
• Notes: Convert various document formats to markdown

microsoftdocs

• File: shared/mcp/microsoft-docs.md
• Type: HTTP (`(learn.microsoft.com/redacted)
• Tools: All tools (documentation search and fetch)
• Secrets: None (public service)
• Notes: Official Microsoft Learn documentation

notion

• File: shared/mcp/notion.md
• Type: Container (mcp/notion)
• Tools: search_pages, get_page, get_database, query_database
• Secrets: NOTION_API_TOKEN
• Notes: Read-only MCP tools + custom safe-job notion-add-comment (requires NOTION_PAGE_ID env var)

semgrep

• File: shared/mcp/semgrep.md
• Type: Container (semgrep/semgrep:latest)
• Tools: All tools (6 tools: semgrep_rule_schema, get_supported_languages, semgrep_scan, semgrep_scan_local, semgrep_scan_with_custom_rule, semgrep_findings)
• Secrets: None
• Notes: Static analysis and security scanning

sentry

• File: shared/mcp/sentry.md
• Type: Stdio (npx @sentry/mcp-server@0.29.0)
• Tools: 14 tools including whoami, find_organizations, find_teams, find_projects, find_releases, get_issue_details, get_trace_details, get_event_attachment, search_events, search_issues, find_dsns, analyze_issue_with_seer, search_docs, get_doc
• Secrets: SENTRY_ACCESS_TOKEN, OPENAI_API_KEY (optional, for search_docs)
• Notes: Error tracking and monitoring integration

server-memory

• File: shared/mcp/server-memory.md
• Type: Container (mcp/memory)
• Tools: store_memory, retrieve_memory, list_memories, delete_memory
• Secrets: None
• Notes: Requires cache-memory: true, persistent key-value storage across workflow runs

skillz

• File: shared/mcp/skillz.md
• Type: Container (intellectronica/skillz:latest)
• Tools: Dynamic (based on .github/skills/ directory)
• Secrets: None (uses github.token)
• Notes: Loads Claude-style skills from repository, mounts .github/skills/ read-only

slack

• File: shared/mcp/slack.md
• Type: Custom safe-job (post-to-slack-channel)
• Tools: N/A (not an MCP server)
• Secrets: SLACK_BOT_TOKEN
• Notes: Posts messages to Slack channels (max 200 chars), requires GH_AW_SLACK_CHANNEL_ID env var, supports staged mode preview

svelte

• File: shared/mcp/svelte.md
• Type: HTTP (`(mcp.svelte.dev/redacted)
• Tools: list-sections, get-documentation, svelte-autofixer, playground-link, svelte_definition
• Secrets: None (public service)
• Notes: Svelte 5 and SvelteKit documentation

tavily

• File: shared/mcp/tavily.md
• Type: HTTP (`(mcp.tavily.com/redacted)
• Tools: All tools (search and research)
• Secrets: TAVILY_API_KEY
• Notes: AI-powered search and research platform

Key Observations

Type Distribution:

• Container-based (11): Most common deployment method, includes ast-grep, azure, brave-search, chroma, context7, markitdown, notion, semgrep, server-memory, skillz, and arXiv
• HTTP (5): Remote services (datadog, deepwiki, microsoftdocs, svelte, tavily)
• Local HTTP (2): drain3 (Python-based), jupyter (Docker Compose services)
• Stdio (2): fabric-rti (uvx), sentry (npx)
• Safe-Job (1): slack (custom GitHub Actions job, not an MCP server)

Authentication Patterns:

• No Secrets Required (11): arxiv, ast-grep, chroma, deepwiki, drain3, jupyter, markitdown, microsoftdocs, semgrep, server-memory, skillz, svelte
• API Keys (6): brave-search, context7, datadog (2 keys), notion, sentry, tavily
• Azure Service Principal (2): azure, fabric-rti
• Bot Token (1): slack

Special Features:

• Cache-Memory Integration (3): chroma (dedicated cache ID), drain3 (state persistence), server-memory (persistent storage)
• Read-Only Mode (2): azure (--read-only flag), fabric-rti (excludes write operations)
• Custom Safe-Jobs (2): slack (post-to-slack-channel), notion (notion-add-comment)
• Dynamic Tools (1): skillz (loads from .github/skills/)
• Public Services (4): deepwiki, microsoftdocs, svelte (no authentication required)

Advanced Configurations:

• drain3: Complex local setup with Python, FastMCP, streaming JSONL, health checks
• jupyter: Multi-service Docker Compose with health checks
• azure: Comprehensive Azure SDK access (40+ services)
• fabric-rti: Microsoft Fabric RTI with KQL query support

Recommendations

1. Consistency: All configurations follow proper frontmatter YAML structure and are well-documented
2. Security: Azure and fabric-rti properly implement read-only modes; secrets are consistently referenced via ${{ secrets.* }}
3. Documentation: Each configuration includes usage examples and setup instructions
4. Status: All 21 servers are properly configured and ready for use ✅

References

• Inspection Data: /tmp/gh-aw/cache-memory/mcp-inspections/2026-02-09.json
• Source Directory: .github/workflows/shared/mcp/

AI generated by MCP Inspector Agent

• expires on Feb 16, 2026, 6:36 PM UTC

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-02-16T18:36:14.429Z.

Closed by Workflow