Skip to content

Top 50 MCP Servers Security Audit & Isolation Benchmark

Density:

Hermes Codex independently benchmarks every mainstream MCP server across four critical dimensions:

  1. Isolation Score (0โ€“100): Evaluates runtime sandboxing, principle of least privilege, memory bounds, and syscall surface.
  2. Blast Radius (Low / Medium / High / Critical): Quantifies potential damages if the serverโ€™s tools are tricked or hijacked by an adversary.
  3. Attack Surface: Identifies specific exploit primitives (arbitrary command execution, SQL injection, SSRF, credential scraping).
  4. Confinement Playbook: Prescribes exact, reproducible hardening directives (Docker micro-containers, seccomp filters, readonly mounts, JIT tokens).

Explore the interactive registry below to audit your agentic dependencies, filter by category or blast radius tier, and export your stack configuration.

๐Ÿ”Œ MCP ECOSYSTEM KNOWLEDGE BASE
HERMES HASS v1.0 AUDIT ๐Ÿ“… 2026-10-08

Top 50 MCP Servers Security Audit & Isolation Benchmark

Comprehensive attack surface evaluation, isolation scores (0-100), kinetic blast radius ratings, and confinement playbooks across the top 50 deployed Model Context Protocol servers.

50 Audited Servers
43 / 100 Mean Isolation Score
19 Critical Blast Radius
27 High Blast Radius
๐Ÿ”
CRITICAL BLAST Storage & Filesystem
28 /100

@modelcontextprotocol/server-filesystem

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ fs_read๐Ÿ”‘ fs_write๐Ÿ”‘ path_traversal_risk
Attack Surface:

Direct filesystem access allowing agents to read and mutate directory structures. Vulnerable to symlink escapes, path traversal, and sensitive config harvesting.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Enforce chroot or Linux container mount namespaces. Run with read-only root mount and explicit whitelist of disposable subdirectories.

Related Patterns & CVEs:
CRITICAL BLAST System & Shell
12 /100

mcp-server-bash

Community Core โ€ข stdio
๐Ÿ”‘ shell_exec๐Ÿ”‘ fs_write๐Ÿ”‘ network_outbound๐Ÿ”‘ env_secrets
Attack Surface:

Arbitrary shell execution on host. Direct RCE blast radius with access to ambient environment variables and process memory.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Mandatory ephemeral MicroVM (Firecracker/gVisor). Disallow host subshell access completely; strip parent env secrets.

Related Patterns & CVEs:
CRITICAL BLAST System & Shell
22 /100

docker-mcp-server

Docker Inc. โ€ข stdio
๐Ÿ”‘ docker_socket๐Ÿ”‘ container_lifecycle๐Ÿ”‘ host_mount
Attack Surface:

Interacts with Docker daemon socket. Unsanitized volume mounts enable immediate root breakout onto the Docker host.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Rootless Docker or Podman only. Prohibit mounting host / or docker.sock inside spawned containers.

Related Patterns & CVEs:
CRITICAL BLAST System & Shell
25 /100

kubernetes-mcp-server

CNCF Community โ€ข stdio
๐Ÿ”‘ k8s_api๐Ÿ”‘ pod_exec๐Ÿ”‘ secret_read๐Ÿ”‘ namespace_mutations
Attack Surface:

Cluster-wide orchestration via kubeconfig. Agent prompt injection can trigger unauthorized pod deployments or secret dumping.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

RBAC scoping to dedicated non-privileged namespace. Disable secret viewing and pod exec verbs.

Related Patterns & CVEs:
CRITICAL BLAST System & Shell
30 /100

terminal-sandbox-mcp

Model Context Protocol โ€ข stdio
๐Ÿ”‘ pty_alloc๐Ÿ”‘ process_spawn๐Ÿ”‘ fs_read
Attack Surface:

Allocates interactive PTY sessions. Threat actors use blind command injection to spawn reverse shells.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Attach strict seccomp filters and drop CAP_SYS_ADMIN. Restrict stdout/stderr buffer lengths.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
38 /100

@modelcontextprotocol/server-postgres

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ db_query๐Ÿ”‘ db_mutation๐Ÿ”‘ schema_read
Attack Surface:

Executes SQL statements on PostgreSQL databases. Susceptible to indirect prompt injection driving second-order SQL injection.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Read-only connection credentials. Block DDL commands (DROP, ALTER, CREATE) and use parameterized statement wrappers.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
45 /100

@modelcontextprotocol/server-sqlite

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ file_access๐Ÿ”‘ db_query๐Ÿ”‘ db_write
Attack Surface:

Local SQLite database operations. ATTACH DATABASE commands allow writing outside intended storage files.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Open in SQLITE_OPEN_READONLY mode unless explicit write transaction required. Disable ATTACH DATABASE.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
40 /100

mysql-mcp-server

Community โ€ข stdio
๐Ÿ”‘ db_query๐Ÿ”‘ db_mutation๐Ÿ”‘ user_privileges
Attack Surface:

MySQL queries and table scans. Threat actors inject SELECT INTO OUTFILE to plant web shells.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Disable FILE privilege. Require human approval for any non-SELECT statements.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
42 /100

redis-mcp-server

Community โ€ข stdio
๐Ÿ”‘ cache_read๐Ÿ”‘ cache_write๐Ÿ”‘ flushall_risk
Attack Surface:

Redis in-memory caching and session store. FlushAll or Config Set commands risk denial-of-service and host file overwrite.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Rename or disable dangerous commands (CONFIG, FLUSHALL, FLUSHDB, EVAL). Namespace agent keys with prefixes.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
44 /100

mongodb-mcp-server

Community โ€ข stdio
๐Ÿ”‘ nosql_query๐Ÿ”‘ document_mutations๐Ÿ”‘ collection_drop
Attack Surface:

MongoDB NoSQL queries. NoSQL operator injection ($where, $regex) allows bypassing filters and exfiltrating collections.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Disallow $where JavaScript evaluation on MongoDB server. Enforce read-only collections.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
48 /100

snowflake-mcp-server

Snowflake Community โ€ข stdio
๐Ÿ”‘ warehouse_query๐Ÿ”‘ stage_read๐Ÿ”‘ data_export
Attack Surface:

Enterprise data warehouse querying. Prompt injection can trigger mass table unloads to untrusted S3 buckets.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Block COPY INTO @external_stage commands. Restrict warehouse credits and query duration timeouts.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
50 /100

bigquery-mcp-server

Google Cloud Community โ€ข stdio
๐Ÿ”‘ bigquery_jobs๐Ÿ”‘ dataset_read๐Ÿ”‘ billing_spend
Attack Surface:

Executes BigQuery analytics. Rogue queries can incur denial-of-wallet costs through petabyte-scale table scans.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Enforce maximum bytes billed quotas per query. Assign BigQuery Data Viewer role without job creation permissions.

Related Patterns & CVEs:
HIGH BLAST Databases & Warehouses
52 /100

neo4j-mcp-server

Neo4j Inc. โ€ข stdio
๐Ÿ”‘ cypher_exec๐Ÿ”‘ graph_mutations๐Ÿ”‘ procedures
Attack Surface:

Cypher queries on knowledge graphs. APOC procedure calls can allow host filesystem reading or network requests.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Disable apoc.import and apoc.export procedures. Run with read-only Cypher endpoint.

Related Patterns & CVEs:
CRITICAL BLAST Developer Tools & Git
35 /100

@modelcontextprotocol/server-github

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ repo_read๐Ÿ”‘ repo_write๐Ÿ”‘ pull_request๐Ÿ”‘ workflow_dispatch
Attack Surface:

GitHub API automation. Allows commits, branch pushes, and CI workflow triggering. Attacker can poison source code or backdoors into master branch.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Strictly prohibit push to default branch. Enforce Branch Protection Rules requiring 2 human approvals for agent-authored PRs.

CRITICAL BLAST Developer Tools & Git
38 /100

gitlab-mcp-server

GitLab Community โ€ข stdio
๐Ÿ”‘ repo_read๐Ÿ”‘ repo_write๐Ÿ”‘ pipeline_trigger๐Ÿ”‘ variables_read
Attack Surface:

GitLab repository & CI pipeline control. Risk of harvesting CI/CD masked variables and secret tokens.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Scope Personal Access Token without admin:api. Prevent reading masked project CI variables.

Related Patterns & CVEs:
HIGH BLAST Developer Tools & Git
42 /100

@modelcontextprotocol/server-git

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ git_cli๐Ÿ”‘ fs_read๐Ÿ”‘ fs_write
Attack Surface:

Local git repository inspections. Submodule configurations and .git/hooks can execute arbitrary shell scripts on checkout.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Disable git hooks via git -c core.hooksPath=/dev/null. Run in isolated git worktree.

Related Patterns & CVEs:
MODERATE BLAST Developer Tools & Git
62 /100

@modelcontextprotocol/server-sentry

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ sentry_api๐Ÿ”‘ stacktrace_read๐Ÿ”‘ issue_update
Attack Surface:

Sentry issue retrieval. Stack traces often contain embedded API secrets, JWT tokens, and user PII.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Scrub server-side sensitive variables in Sentry settings before passing issues to agent context.

Related Patterns & CVEs:
MODERATE BLAST Developer Tools & Git
60 /100

datadog-mcp-server

Datadog Community โ€ข stdio
๐Ÿ”‘ metric_query๐Ÿ”‘ log_search๐Ÿ”‘ apm_traces
Attack Surface:

Datadog observability queries. Log searches can expose plain-text passwords or internal infrastructure topology.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Filter sensitive log indexes. Limit API key permissions to metrics and dashboard reading.

Related Patterns & CVEs:
HIGH BLAST Web & Browsing
65 /100

@modelcontextprotocol/server-brave-search

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ web_search๐Ÿ”‘ api_token
Attack Surface:

Web search engine query execution. Primary conduit for indirect prompt injection via untrusted search snippets.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Sanitize returned snippets through dual-channel LLM verifier before blending into agent system memory.

Related Patterns & CVEs:
CRITICAL BLAST Web & Browsing
48 /100

@modelcontextprotocol/server-fetch

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ http_client๐Ÿ”‘ ssrf_risk๐Ÿ”‘ redirect_follow
Attack Surface:

Generic HTTP GET/POST fetch tool. High SSRF risk targeting cloud metadata services (169.254.169.254) and internal networks.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Block RFC 1918 private IP ranges and cloud metadata endpoints (169.254.169.254) at network and DNS levels.

Related Patterns & CVEs:
CRITICAL BLAST Web & Browsing
32 /100

@modelcontextprotocol/server-puppeteer

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ chromium_exec๐Ÿ”‘ js_eval๐Ÿ”‘ screenshot๐Ÿ”‘ cookies_read
Attack Surface:

Headless Chromium browser automation. Malicious web pages can trigger zero-day sandbox escapes or steal session cookies.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Launch Chrome with --no-sandbox forbidden in production; run within disposable MicroVM with isolated network interface.

Related Patterns & CVEs:
CRITICAL BLAST Web & Browsing
34 /100

playwright-mcp-server

Community โ€ข stdio
๐Ÿ”‘ browser_context๐Ÿ”‘ file_download๐Ÿ”‘ network_proxy
Attack Surface:

Multi-browser web testing automation. Similar to Puppeteer, browser contexts can be subverted via untrusted JavaScript payloads.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Disable file downloads. Restrict navigation to approved domains via DNS proxy filtering.

Related Patterns & CVEs:
HIGH BLAST Web & Browsing
68 /100

tavily-search-mcp

Tavily AI โ€ข stdio
๐Ÿ”‘ search_api๐Ÿ”‘ summary_read
Attack Surface:

AI-optimized web search. Susceptible to search result adversarial poisoning targeting agent decision trees.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Validate parsed search entities against internal ground truth before performing operational actions.

Related Patterns & CVEs:
HIGH BLAST Productivity & Collab
45 /100

@modelcontextprotocol/server-slack

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ slack_channels_read๐Ÿ”‘ slack_messages_write๐Ÿ”‘ user_impersonation
Attack Surface:

Slack bot interaction. Adversaries posting messages in monitored channels can inject indirect prompt instructions to hijack agent.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Restrict agent bot to dedicated testing channel. Never allow posting with user-impersonation tokens.

Related Patterns & CVEs:
HIGH BLAST Productivity & Collab
48 /100

@modelcontextprotocol/server-gdrive

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ gdrive_read๐Ÿ”‘ gdrive_search๐Ÿ”‘ oauth_token
Attack Surface:

Google Drive file search and download. Poisoned shared Google Docs can inject persistent backdoor instructions.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Strictly isolate corporate drives by department. Use OAuth service accounts with read-only view scopes.

Related Patterns & CVEs:
HIGH BLAST Productivity & Collab
52 /100

notion-mcp-server

Community โ€ข stdio
๐Ÿ”‘ notion_blocks_read๐Ÿ”‘ notion_pages_write๐Ÿ”‘ api_token
Attack Surface:

Notion workspace page and database sync. Injection inside Notion blocks can trigger unauthorized exfiltration.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Scope Notion integration token to designated sub-pages only. Prohibit editing workspace settings.

Related Patterns & CVEs:
MODERATE BLAST Productivity & Collab
55 /100

linear-mcp-server

Linear Community โ€ข stdio
๐Ÿ”‘ issues_read๐Ÿ”‘ issues_write๐Ÿ”‘ project_status
Attack Surface:

Linear task tracking. Malicious issue descriptions can manipulate agent prioritizations and sprint planning.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Limit API token to read-only on sensitive projects. Disallow deleting issues.

Related Patterns & CVEs:
HIGH BLAST Productivity & Collab
46 /100

sooperset-mcp-atlassian

Atlassian Community โ€ข stdio
๐Ÿ”‘ jira_api๐Ÿ”‘ confluence_api๐Ÿ”‘ project_admin
Attack Surface:

Jira and Confluence corporate knowledge integration. Widespread enterprise blast radius across internal tickets and documentation.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Enforce principle of least privilege on Atlassian API tokens. Restrict access to internal security boards.

Related Patterns & CVEs:
HIGH BLAST Productivity & Collab
50 /100

discord-mcp-server

Community โ€ข stdio
๐Ÿ”‘ discord_bot๐Ÿ”‘ guild_messages๐Ÿ”‘ user_dm
Attack Surface:

Discord community bot operations. Untrusted user DMs can hijack customer-facing agents and broadcast phishing links.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Sanitize Discord user input with strict content moderation before agent processing. Block sending external URLs.

Related Patterns & CVEs:
CRITICAL BLAST Productivity & Collab
42 /100

email-imap-mcp-server

Community โ€ข stdio
๐Ÿ”‘ mailbox_read๐Ÿ”‘ attachments_extract๐Ÿ”‘ credentials_stored
Attack Surface:

IMAP mailbox ingestion. Inbound phishing emails with hidden invisible text prompt injections can seize full control of the agent.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Strip HTML comments, invisible zero-width unicode, and CSS display:none text prior to LLM reasoning.

Related Patterns & CVEs:
CRITICAL BLAST Productivity & Collab
44 /100

email-smtp-mcp-server

Community โ€ข stdio
๐Ÿ”‘ smtp_send๐Ÿ”‘ domain_spoofing๐Ÿ”‘ mass_mail
Attack Surface:

Outbound SMTP email dispatch. Hijacked agents can be utilized as spam relays or executive BEC impersonation channels.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Require out-of-band cryptographic approval for any email dispatched to external recipient domains.

Related Patterns & CVEs:
HIGH BLAST Agentic Memory & RAG
48 /100

@modelcontextprotocol/server-memory

Anthropic / MCP Core โ€ข stdio
๐Ÿ”‘ graph_memory_read๐Ÿ”‘ graph_memory_write๐Ÿ”‘ context_poisoning
Attack Surface:

Persistent knowledge graph memory. Poisoned memories persist across sessions, establishing durable agent backdoors.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Implement periodic memory auditing and provenance signatures on every stored entity and relation.

Related Patterns & CVEs:
HIGH BLAST Agentic Memory & RAG
54 /100

pinecone-mcp-server

Pinecone Systems โ€ข stdio
๐Ÿ”‘ vector_upsert๐Ÿ”‘ vector_query๐Ÿ”‘ namespace_mutations
Attack Surface:

Vector database operations. Attacker injects adversarial embeddings to hijack retrieval-augmented generation (RAG).

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Namespace vector spaces by tenant and verify cosine anomaly thresholds on newly upserted vectors.

Related Patterns & CVEs:
HIGH BLAST Agentic Memory & RAG
56 /100

qdrant-mcp-server

Qdrant Community โ€ข stdio
๐Ÿ”‘ collection_upsert๐Ÿ”‘ payload_filtering๐Ÿ”‘ snapshot_export
Attack Surface:

Qdrant vector search. Payload tampering can manipulate filter logic and retrieve unauthorized records.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Sign and validate payload metadata with tamper-evident HMAC keys.

Related Patterns & CVEs:
HIGH BLAST Agentic Memory & RAG
55 /100

weaviate-mcp-server

Weaviate Community โ€ข stdio
๐Ÿ”‘ graphql_query๐Ÿ”‘ schema_mutations๐Ÿ”‘ object_insert
Attack Surface:

Weaviate vector store with GraphQL capabilities. Schema injection allows modifying class definitions.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Disable schema mutation APIs during production runtime.

Related Patterns & CVEs:
MODERATE BLAST Agentic Memory & RAG
58 /100

chroma-mcp-server

Chroma Community โ€ข stdio
๐Ÿ”‘ collection_query๐Ÿ”‘ document_embedding๐Ÿ”‘ local_storage
Attack Surface:

Chroma embedding database. Local SQLite-backed deployments can suffer file locking or local database corruption.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Enforce client-side cryptographic hashing on text documents before embedding.

Related Patterns & CVEs:
CRITICAL BLAST Cloud & IAM
24 /100

amazon-aws-mcp-server

Amazon Web Services โ€ข stdio
๐Ÿ”‘ aws_sdk๐Ÿ”‘ s3_buckets๐Ÿ”‘ iam_sts๐Ÿ”‘ ec2_instances
Attack Surface:

Broad AWS management capability. Hijacked agents can deploy crypto-miners or delete corporate S3 backups.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Enforce tight AWS IAM Service Control Policies (SCPs) and require MFA token generation for mutating actions.

Related Patterns & CVEs:
CRITICAL BLAST Cloud & IAM
26 /100

google-cloud-mcp-server

Google Cloud Community โ€ข stdio
๐Ÿ”‘ gcp_iam๐Ÿ”‘ gcs_buckets๐Ÿ”‘ cloud_compute๐Ÿ”‘ cloud_functions
Attack Surface:

Google Cloud Platform management. Risk of IAM privilege escalation via service account token creation.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Restrict iam.serviceAccountTokenCreator permissions. Use short-lived OAuth access tokens with 5-minute TTL.

Related Patterns & CVEs:
CRITICAL BLAST Cloud & IAM
25 /100

azure-mcp-server

Microsoft Azure Community โ€ข stdio
๐Ÿ”‘ azure_arm๐Ÿ”‘ entra_id๐Ÿ”‘ blob_storage๐Ÿ”‘ key_vault
Attack Surface:

Azure Resource Manager & Entra ID control. Agent compromise exposes cloud tenant directory and key vault secrets.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Lock down Key Vault access policies. Enforce Azure Privileged Identity Management (PIM) with approval workflows.

Related Patterns & CVEs:
CRITICAL BLAST Cloud & IAM
36 /100

cloudflare-mcp-server

Cloudflare Community โ€ข stdio
๐Ÿ”‘ dns_records๐Ÿ”‘ workers_deploy๐Ÿ”‘ kv_storage๐Ÿ”‘ waf_rules
Attack Surface:

Cloudflare edge management. Attacker can subvert DNS records or disable WAF rules to expose origin servers.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Separate DNS administration token from CDN caching token. Restrict zone DNS modification rights.

Related Patterns & CVEs:
HIGH BLAST CRM & Billing
46 /100

salesforce-mcp-server

Salesforce Community โ€ข stdio
๐Ÿ”‘ soql_query๐Ÿ”‘ lead_export๐Ÿ”‘ contact_mutation
Attack Surface:

Salesforce enterprise CRM operations. Direct exposure to confidential sales pipelines, customer contracts, and lead records.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Restrict SOQL queries to active operational fields. Mask all customer credit card and banking fields.

Related Patterns & CVEs:
HIGH BLAST CRM & Billing
50 /100

hubspot-mcp-server

HubSpot Community โ€ข stdio
๐Ÿ”‘ contacts_read๐Ÿ”‘ campaigns_send๐Ÿ”‘ deals_update
Attack Surface:

HubSpot marketing and contact automation. Risk of unauthorized mass marketing dispatch or PII database dump.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Scope private app token to read-only contacts. Disallow email marketing batch API dispatch.

Related Patterns & CVEs:
HIGH BLAST CRM & Billing
48 /100

zendesk-mcp-server

Zendesk Community โ€ข stdio
๐Ÿ”‘ tickets_read๐Ÿ”‘ tickets_reply๐Ÿ”‘ user_records
Attack Surface:

Zendesk customer support ticketing. Attacker can read internal escalations or post malicious phishing replies.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Enforce human review on public replies. Redact customer passwords and auth tokens automatically.

Related Patterns & CVEs:
CRITICAL BLAST CRM & Billing
30 /100

stripe-mcp-server

Community โ€ข stdio
๐Ÿ”‘ charges_create๐Ÿ”‘ refunds_issue๐Ÿ”‘ customers_read
Attack Surface:

Stripe payments and financial transactions. Severe kinetic blast radius: rogue refund issuances or unauthorized payments.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Use restricted API keys with read-only charges. Require physical hardware token confirmation for any transfer.

Related Patterns & CVEs:
CRITICAL BLAST AI Platforms & Hubs
18 /100

obot-mcp-catalog-server

Obot Platform โ€ข stdio
๐Ÿ”‘ tool_discovery๐Ÿ”‘ catalog_credentials๐Ÿ”‘ agent_delegation
Attack Surface:

Obot enterprise tool catalog. Directly impacted by CVE-2026-105138 allowing regular users to dump static tool secrets.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Upgrade to Obot >= 0.26.2 immediately. Enforce RBAC on MCP catalog endpoints and migrate to ephemeral tokens.

HIGH BLAST AI Platforms & Hubs
52 /100

ollama-mcp-server

Ollama Community โ€ข stdio
๐Ÿ”‘ model_pull๐Ÿ”‘ model_generate๐Ÿ”‘ system_memory
Attack Surface:

Local model serving and execution. Model pulling can download poisoned GGUF weights with embedded exploits.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Verify SHA256 checksums of model weights against verified vendor manifests before loading into memory.

Related Patterns & CVEs:
HIGH BLAST AI Platforms & Hubs
55 /100

huggingface-mcp-server

Hugging Face Community โ€ข stdio
๐Ÿ”‘ dataset_download๐Ÿ”‘ model_inference๐Ÿ”‘ api_token
Attack Surface:

Hugging Face Hub interaction. Loading untrusted pickle-based models causes arbitrary Python code execution.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Enforce safetensors format only; block loading PyTorch pickle files completely.

Related Patterns & CVEs:
HIGH BLAST Productivity & Collab
48 /100

telegram-mcp-server

Community โ€ข stdio
๐Ÿ”‘ telegram_bot๐Ÿ”‘ channel_post๐Ÿ”‘ file_download
Attack Surface:

Telegram messaging bot operations. Subject to IPC record injection and malicious attachment harvesting.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Sanitize incoming chat messages and block automatic attachment download without manual inspection.

Related Patterns & CVEs:
CRITICAL BLAST System & Shell
20 /100

tmux-mcp-server

Community โ€ข stdio
๐Ÿ”‘ tmux_sessions๐Ÿ”‘ pane_injection๐Ÿ”‘ shell_access
Attack Surface:

Tmux session manager. Injects keystrokes directly into active terminal panes, bypassing shell argument filters.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Prohibit attaching to existing human user sessions. Run within dedicated ephemeral pty container.

Related Patterns & CVEs:
HIGH BLAST Productivity & Collab
52 /100

obsidian-mcp-server

Community โ€ข stdio
๐Ÿ”‘ vault_read๐Ÿ”‘ vault_write๐Ÿ”‘ markdown_links
Attack Surface:

Obsidian markdown note vault access. Threat actors inject hidden markdown links and exfiltration payloads into daily notes.

๐Ÿ›ก๏ธ Hermes Confinement Playbook:

Confine vault access to a sandboxed read-only clone. Strip external URI schemes from generated notes.

Related Patterns & CVEs:

MCP Security Architecture Guide

Read the foundational guide on hardening Anthropicโ€™s Model Context Protocol against tool poisoning and rogue servers. Explore MCP Security Guide โ†’

Declare Your Security Stack (MyStack)

Declare your deployed AI frameworks, inference backends, and MCP tools to evaluate real-time exposure. Open MyStack Declaration โ†’

Agentic Attack Patterns

Analyze verified adversarial patterns including function hijacking, semantic rug-pulls, and agent poisoning. Browse Attack Patterns โ†’