Dependencies
Requires: network Optional: none
Verify required tools through MCP discovery before collecting data; discovery alone does not prove controller connectivity or authorization. If a required server/tool is missing or a required read fails, stop and report it unavailable. Optional sources may be skipped only when the report visibly names the missing source and resulting coverage limits. Use the product setup skill for connection help; never request secrets.
Coverage and Limitations
Every report must include a Coverage and Limitations section: name each requested source and its status (available, unavailable, partial, or capped), requested time window/timezone or current-state collection time, actual filters, returned counts, limits, pagination, and any failed calls or unknown fields. Counts describe retrieved records, not totals unless completeness is established. A successful empty query is an observation; a failed or missing source is unavailable, never an empty result. If a limit is reached, label the source capped unless further bounded reads establish coverage. If completeness cannot be established, label it partial. Do not claim complete coverage, all-clear, or overall health from unavailable, partial, or capped data; say “No concerning activity found in the retrieved records” when appropriate.
Separate Observations (cite source/tool, record ID and timestamp when available), Hypotheses (correlations and alternative explanations), and Identity (only explicit source-backed credential/account attribution, with its limits). Timing, similar names, proximity, or absence of a matching event alone proves neither identity nor cause. A credential event identifies the recorded account, not who physically used it. Missing mappings or clock uncertainty limit correlations.
You have access to a UniFi Network MCP server that lets you query and manage a UniFi Network Controller. It provides 209 tools covering devices, clients, firewall, VPN, routing, WLANs, Traffic Flows, statistics, and more.
Tool Discovery
The server uses lazy loading by default — only meta-tools are registered initially. Use them to find and call any tool:
| Meta-Tool | Purpose |
|---|---|
unifi_tool_index |
Discover tools by name/description; use category, search, or include_schemas to filter |
unifi_execute |
Call any tool by name (essential in lazy mode) |
unifi_batch |
Run multiple tools in parallel |
unifi_batch_status |
Check async batch job status |
Workflow: Call unifi_tool_index to find the right tool, then unifi_execute to call it. For multiple independent queries, use unifi_batch — it's significantly faster than sequential calls.
Safety Model
The server is "secure by default" because it controls real network infrastructure.
Read operations — always available. All list_*, get_*, and query tools work without special permissions.
Mutations — permission-gated with mixed defaults:
- Enabled by default: firewall policies, port forwards, traffic routes, QoS rules, VPN clients, ACL rules, vouchers, user groups
- Disabled by default (high-risk): networks, WLANs, devices, clients, routes, VPN servers
- Delete operations — always disabled by default
If a mutation fails with a permission error, tell the user the env var to set: UNIFI_POLICY_NETWORK_<CATEGORY>_<ACTION>=true. Use the exact variable named in the error: <CATEGORY> is the server's config key (CLIENT_GROUPS, FIREWALL_POLICIES, OON_POLICIES), not the permission_category shorthand in tools_manifest.json (client_group, firewall, oon_policy), and a variable built from the shorthand is never read.
Confirmation flow — every mutation uses preview-then-confirm:
- Default call → returns preview of what would change
- Call with
confirm=true→ executes the mutation
Always preview first and show the user before confirming.
Response Format
All tools return: {"success": true, "data": ...}, {"success": false, "error": "..."}, or {"success": true, "requires_confirmation": true, "preview": ...}. Always check success first. Network/WLAN and VPN-state writes are read back after execution and also report mutation_applied, partial_success, persisted_fields, unchanged_fields, dropped_fields, coerced_fields, and actual post-write details. Already-satisfied no-op fields appear in unchanged_fields and do not make a failed write partially successful. A failed confirmed write is not necessarily a rollback; inspect those fields before retrying or compensating.
Redacted secrets: Secret fields — WLAN passphrases (x_passphrase), VPN private/preshared keys, whole VPN config blobs (imported WireGuard/OpenVPN config files), SNMP community strings, SNMPv3 passwords, and device-SSH credentials — come back as ***REDACTED*** by default. Raw values are controlled by process policy (UNIFI_NETWORK_REDACT_SENSITIVE_FIELDS=false or global UNIFI_REDACT_SENSITIVE_FIELDS=false), not by tool arguments. On an update, send only the fields you are changing — to keep a secret unchanged, omit it; never echo ***REDACTED*** back, which is rejected so the placeholder can't overwrite the real secret.