Infrahub Sot

MCP Tools (10 tools)

Node Operations (3 tools)

Schema Operations (3 tools)

GraphQL Operations (2 tools)

Branch Operations (2 tools)

Workflow: Discover Available Data

When first connecting to Infrahub:

1. List kinds: get_schema_mapping — what infrastructure types are modeled?
2. Inspect schema: get_schema(kind="InfraDevice") — what attributes and relationships does a device have?
3. List filters: get_node_filters(kind="InfraDevice") — how can I query devices?
4. Get nodes: get_nodes(kind="InfraDevice") — list all devices
5. Report: infrastructure data model overview with node counts per kind

Workflow: Infrastructure Audit

When auditing infrastructure state in Infrahub:

1. Schema overview: get_schema_mapping — discover all kinds
2. Device inventory: get_nodes(kind="InfraDevice") — all devices
3. IP addresses: get_nodes(kind="InfraIPAddress") — all IPs (if IPAM is modeled)
4. Prefixes: get_nodes(kind="InfraPrefix") — all subnets
5. Relationships: get_related_nodes(kind="InfraDevice", relation="interfaces") — device interfaces
6. Report: infrastructure inventory from Infrahub with relationship context

Workflow: Branch-Based Change

When proposing an infrastructure change:

1. List branches: get_branches — see existing branches
2. Create branch: branch_create(name="change-123-add-vlan") — isolate changes
3. Query current state: get_nodes(kind="InfraVLAN", branch="change-123-add-vlan") — view on branch
4. Make changes via GraphQL: query_graphql(query="mutation { ... }", branch="change-123-add-vlan")
5. Verify: get_nodes on branch — confirm changes look correct
6. Merge: (via Infrahub UI/API — merge branch to main)
7. Report: change summary with before/after branch comparison

Workflow: GraphQL Exploration

When building custom queries:

1. Schema: get_graphql_schema — full SDL schema, understand query structure
2. Test query: query_graphql(query="{ InfraDevice { edges { node { name { value } } } } }")
3. Filtered query: query_graphql(query="{ InfraDevice(name__value: \"core-rtr\") { ... } }")
4. Nested relationships: traverse in a single query via GraphQL nesting
5. Report: custom data extraction with exactly the fields needed

Integration with Other Skills

Infrahub vs NetBox vs Nautobot

NetClaw supports all three source-of-truth platforms:

Important Rules

• Discover before querying — always call get_schema_mapping first to learn what kinds exist. Don't guess kind names.
• Use filters — call get_node_filters to learn valid filter syntax before using get_nodes with filters.
• Branch for changes — create a branch before making mutations via query_graphql. Never mutate main directly.
• GraphQL mutations require authorization — ensure the API token has write permissions for mutation queries.
• Partial matching — use partial_match=True in get_nodes for fuzzy matching on filter values.
• Relationship traversal — use get_related_nodes to follow relationships; use get_schema to discover relationship names first.
• Record in GAIT — log all Infrahub queries, branch operations, and infrastructure changes.

Environment Variables

• INFRAHUB_ADDRESS — Infrahub instance URL (e.g., http://infrahub.example.com:8000)
• INFRAHUB_API_TOKEN — Infrahub API authentication token
• MCP_HOST — Server bind address when running in HTTP mode (default: 0.0.0.0, optional)
• MCP_PORT — Server port when running in HTTP mode (default: 8001, optional)