About Anomalyarmor
AnomalyArmor is an MCP server for engineering teams to inspect and manage warehouse observability through natural-language interactions. It covers alerts, freshness, schema drift, validity checks, and lineage, with remote OAuth or local API-key connection options.
Key features
- 01Query and manage alerts and rules
- 02Monitor freshness and schema changes
- 03Configure metrics, validity, and referential checks
- 04Inspect data lineage and tags
- 05Track asynchronous analysis jobs
Use cases
- 01Investigate unresolved alerts and data trends
- 02Find stale tables or schema drift
- 03Configure quality rules and review results
- 04Trace upstream dependencies and job status
MCP support
The server offers remote MCP access over Streamable HTTP and can also run locally with an API key. Its structured tools cover warehouse health, monitoring configuration, data quality, lineage, and job management.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.anomalyarmor.ai/mcpclaude mcp add --transport http agents https://mcp.anomalyarmor.ai/mcp{"mcpServers":{"agents":{"type":"http","url":"https://mcp.anomalyarmor.ai/mcp"}}}Open in Cursor{"servers":{"agents":{"type":"http","url":"https://mcp.anomalyarmor.ai/mcp"}}}Open in VS Code{"mcpServers":{"agents":{"type":"http","url":"https://mcp.anomalyarmor.ai/mcp"}}}{"mcpServers":{"agents":{"serverUrl":"https://mcp.anomalyarmor.ai/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project READMEarmor-mcp
The AnomalyArmor MCP server lets AI assistants (Claude Code, Cursor, Claude Desktop, any MCP client) interact with your data warehouse through 50+ structured tools: query alerts, monitor freshness, investigate schema drift, configure validity rules, and trace lineage in natural language.
- Homepage: https://www.anomalyarmor.ai
- Docs: https://docs.anomalyarmor.ai/integrations/mcp-server
- Server card: https://www.anomalyarmor.ai/.well-known/mcp/server-card.json
- Status: Production. Versioned MCP protocol
2025-06-18. OAuth 2.1 + API key auth.
Two ways to connect
| Method | Auth | Best for |
|---|---|---|
| Remote (recommended) | OAuth 2.1 (Clerk) | Zero install, always up to date |
| Local | API key | Air-gapped, on-call laptops, custom proxies |
Remote (recommended)
The server is hosted at https://mcp.anomalyarmor.ai/mcp (Streamable HTTP transport). First call opens a browser for OAuth; the token is cached for ~12h.
Claude Code
claude mcp add anomalyarmor --transport http https://mcp.anomalyarmor.ai/mcp
Cursor
Add to ~/Library/Application Support/Cursor/mcp.json (macOS), ~/.config/Cursor/mcp.json (Linux), or %APPDATA%\Cursor\mcp.json (Windows):
{
"mcpServers": {
"anomalyarmor": {
"url": "https://mcp.anomalyarmor.ai/mcp"
}
}
}
Any MCP client
URL: https://mcp.anomalyarmor.ai/mcp
Transport: streamable-http
Auth: OAuth 2.1
Discovery: https://mcp.anomalyarmor.ai/.well-known/oauth-protected-resource
Local (API key)
Use when the remote server isn't reachable or when you want no third-party dependency.
Install
# Recommended: uvx (no install)
uvx armor-mcp
# Or pip
pip install armor-mcp
Configure
Get an API key from Settings → API Keys at https://app.anomalyarmor.ai.
export ARMOR_API_KEY=aa_live_your_key_here
Or persist to ~/.armor/config.yaml:
api_key: aa_live_your_key_here
Wire to a client
{
"mcpServers": {
"anomalyarmor": {
"command": "uvx",
"args": ["armor-mcp"],
"env": {
"ARMOR_API_KEY": "aa_live_your_key_here"
}
}
}
}
Tools
53 tools across 8 categories. Every tool ships with MCP ToolAnnotations (readOnlyHint, destructiveHint, idempotentHint, openWorldHint: false) so hosts can decide when to auto-confirm.
Health & Briefings
| Tool | Description |
|---|---|
health_summary | Overall warehouse health: stale assets, drift, alerts |
get_todays_briefing | Daily digest with key insights |
get_coverage | Monitoring coverage analysis |
manage_coverage | Update coverage targets |
recommend | AI-suggested next monitoring actions |
Alerts & Incidents
| Tool | Description |
|---|---|
list_alerts | Query alerts with filters (severity, date, status) |
list_inbox_alerts | Alerts not yet triaged |
get_alerts_summary | Aggregate counts and trends |
get_alert_trends | Period-over-period trend analysis |
get_alert_history | Activity timeline for an alert |
update_alert | Acknowledge / resolve / dismiss / snooze |
list_alert_rules | View configured rules |
create_alert_rule | Add a new alert rule |
manage_alert_rule | Update / delete / preview rules |
manage_rule_destinations | Wire rules to destinations |
Assets
| Tool | Description |
|---|---|
list_assets | List data sources with filters |
create_asset | Connect a new source |
manage_asset | Update / delete / test connection |
trigger_asset_discovery | Start schema discovery (async) |
Freshness
| Tool | Description |
|---|---|
get_freshness_summary | Freshness overview |
check_freshness | Check a specific asset / table |
setup_freshness | Bulk-create freshness schedules |
list_freshness_schedules | View configured schedules |
manage_freshness_schedule | Update / delete a schedule |
Schema Monitoring
| Tool | Description |
|---|---|
get_schema_summary | Drift overview |
list_schema_changes | Recent changes with severity |
create_schema_baseline | Capture current schema as a baseline |
enable_schema_monitoring | Start drift detection |
disable_schema_monitoring | Stop drift detection |
Data Quality (Metrics & Validity)
| Tool | Description |
|---|---|
get_metrics_summary | Metric health by asset |
list_metrics | List configured metrics |
create_metric | Add a metric (row count, null %, etc.) |
manage_metric | Update / delete / capture a metric |
get_validity_summary | Pass/fail rate for validity rules |
list_validity_rules | List validity rules |
create_validity_rule | Add a NOT_NULL / UNIQUE / RANGE / REGEX rule |
manage_validity_rule | Update / delete / run a rule |
create_referential_check | Cross-table referential check |
manage_referential | Update / delete a referential check |
Destinations & Routing
| Tool | Description |
|---|---|
list_destinations | Slack / email / webhook / PagerDuty configs |
setup_destination | Add a new destination |
manage_destination | Update / delete / test |
Intelligence, Lineage, Tags, Jobs
| Tool | Description |
|---|---|
ask_question | Synchronous natural-language Q&A about your data |
generate_intelligence | Trigger AI analysis (async, expensive) |
get_lineage | Upstream/downstream dependency graph |
list_tags | Tags for an asset |
create_tag | Tag a table or column |
apply_tags | Bulk tag application |
job_status | Track an async job |
cancel_job | Cancel a running async job |
get_api_key_info | Inspect the current API key's scope |
Full descriptions, parameter schemas, and example prompts: https://docs.anomalyarmor.ai/integrations/mcp-server.
Tool safety conventions
Hosts respect the annotations:
| Category | Pattern | Confirmation |
|---|---|---|
| Read-only | list_*, get_*, check_*, search_* | Auto-invoke |
| Mutating | create_*, update_*, enable_*, disable_*, manage_* | Confirm with user |
| Destructive | delete_* (via manage_*), dismiss_alert | Always confirm; list affected entities |
| Expensive | generate_intelligence, trigger_asset_discovery | Confirm; these queue paid jobs |
Multi-tenancy
The OAuth token (or API key) is scoped to the user's company. You cannot impersonate another tenant. Don't fabricate company_id arguments.
Async jobs
trigger_asset_discovery, generate_intelligence, and metric captures return a job_id. Poll with job_status(job_id) rather than re-triggering.
Discovery surfaces
| File | Purpose |
|---|---|
https://mcp.anomalyarmor.ai/.well-known/oauth-protected-resource | RFC 9728 protected-resource metadata |
https://www.anomalyarmor.ai/.well-known/mcp/server-card.json | SEP-1649 MCP server card |
https://www.anomalyarmor.ai/.well-known/agent-card.json | A2A v0.2 agent card |
https://www.anomalyarmor.ai/.well-known/api-catalog | RFC 9727 API catalog linkset |
Development
git clone https://github.com/anomalyarmor/agents.git
cd agents/armor-mcp
uv sync
uv run pytest
License
MIT — see LICENSE.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.