Give your AI assistant access to every device on your network

latticio is an MCP server that connects Claude, Cursor, and VS Code to your network infrastructure. 345 tools across 7 vendor platforms. Query, diagnose, and configure through natural language.

Explore the Tools See the Features
Coming Soon

Public release is on the way

latticio is in active development. pip install latticio, Docker images, and full documentation will be available at launch. Follow @latticio for updates.

Arista EOS
Cisco IOS-XE
Cisco NX-OS
Juniper JunOS
SONiC
Palo Alto PAN-OS
Fortinet FortiOS
345
MCP Tools
7
Vendors
9
Integrations
11,470+
Tests
92%+
Coverage

How it works

You ask a question. Your AI assistant picks the right tools, queries your devices, and explains what it found.

You: "Are any BGP neighbors down on spine-01?"
  |
AI Assistant (Claude Desktop, Cursor, VS Code)
  |  selects: net_get_bgp_summary(host="spine-01")
  v
latticio → validates input, checks rate limits, selects driver
  |           → queries device via eAPI / RESTCONF / NETCONF / REST
  |           → sanitizes credentials, logs audit trail
  v
AI: "All 4 BGP neighbors on spine-01 are Established.
     leaf-01 and leaf-02 (AS 65001): 24 prefixes each.
     leaf-03 and leaf-04 (AS 65002): 20 prefixes each."

345 tools for everything you need

From basic show commands to full fabric validation, async operations, and AI cluster health monitoring.

Device Info
12 tools — version, model, inventory, health
Interfaces
13 tools — status, counters, errors, optics
Routing
21 tools — BGP, OSPF, routes, ARP/NDP
Switching
14 tools — MLAG, STP, MAC, LLDP, LAG
VLANs
7 tools — list, create, delete, assign
Config Mgmt
10 tools — diffs, save, push, sessions
Security
15 tools — ACLs, AAA, CoPP, SNMP
Monitoring
8 tools — CPU, memory, temp, fans, NTP
Troubleshooting
10 tools — ping, trace, flaps, diagnosis
Fabric Health
6 tools — multi-device validation
EVPN/VXLAN
9 tools — VTEPs, VNIs, EVPN routes
Compliance
5 tools — CIS benchmarks, ANTA
gNMI
5 tools — streaming telemetry
AI/ML Fabric
10 tools — RoCEv2, ECN, PFC, GPU cluster health
Async Tasks
3 tools — background ops with progress tracking
Digital Twin
2 tools — pre-deployment validation via Containerlab
Trending
4 tools — historical metrics analysis
Drift Detection
3 tools — golden config comparison

One interface, seven platforms

All drivers implement the same protocol with 36 normalized getters. Write tools once, run them anywhere.

VendorPlatformProtocolStatus
AristaEOSeAPI (HTTPS)Supported
CiscoIOS-XERESTCONFPreview
CiscoNX-OSNX-APIPreview
JuniperJunOSNETCONFPreview
SONiCSONiCREST APIPreview
Palo AltoPAN-OSXML/REST APIExperimental
FortinetFortiOSREST APIExperimental

Built for production networks

Multiple safety layers active by default. Enterprise auth, RBAC, and MCP protocol compliance for team deployments.

Read-only by default

All write operations blocked unless explicitly enabled. Dangerous commands like reload, write erase, and bash are permanently denied. Configurable deny-list for site-specific needs.

CLI injection prevention

Framework-level protection blocks semicolons, backticks, subshell syntax, and null bytes before any command reaches a device. Tool authors cannot bypass this.

Human-in-the-loop

MCP Elicitation support prompts for user confirmation before destructive operations. Cancel a write before it happens. Graceful fallback when clients don't support it.

Async operations

MCP Tasks primitive for long-running operations. Fabric health checks, compliance audits, and fleet commands return a task handle with progress tracking.

OAuth 2.1 & RBAC

RFC 9728/8707 compliant authentication. Scope-based access control with hierarchy: admin implies write implies read. Step-up authorization for privilege escalation.

Credential sanitization

20+ regex patterns plus keyword-based fallback redact passwords, secrets, community strings, and API keys from every response before reaching the AI assistant.

TLS certificate pinning

Optional per-device certificate pinning for high-security environments. Supports both hostname and IP-addressed connections.

Progressive discovery

Only meta-tools load at startup. Tool categories are loaded on demand as the AI needs them, keeping the context window efficient. Code Mode reduces tokens by 99%.

Audit & observability

Structured JSON audit log with HMAC signing, configurable rotation and retention, syslog forwarding. Prometheus metrics. Per-device circuit breakers and rate limiting.

Semantic validation

Vendor-aware validation catches bad interface names, invalid BGP ASNs, and mismatched parameters before they reach the device. Clear error messages with remediation hints.

Digital twin testing

Validate configuration changes against a Containerlab digital twin before applying to production. Compare live topology to lab for drift detection.

Plugin system

Extend with custom drivers, compliance packs, and tool modules via Python entry points. No core code changes needed.

Plugs into your existing workflow

Inventory, secrets, incident management, metrics, and infrastructure-as-code.

NetBox
Bidirectional DCIM sync, drift detection, cable audits
Nautobot
Inventory source, device context enrichment
ServiceNow
Incident CRUD, CMDB sync, change request validation
HashiCorp Vault
Credential retrieval with token, AppRole, and K8s auth
Prometheus
Metrics export and Grafana dashboard generation
Ansible
Playbook execution, inventory generation
Terraform
State parsing, infrastructure drift detection
Containerlab
Lab lifecycle, digital twin validation, topology comparison
Config Store
Git-backed config versioning with diff and history

Works with any MCP client

Connect latticio to the AI tools you already use.

Claude Desktop
Claude Code
Cursor
VS Code Copilot
ChatGPT Desktop
Any MCP client

Deploy anywhere

From a local subprocess to a Kubernetes cluster with Helm, health checks, and horizontal scaling.

stdio
Default — AI client launches it automatically
Streamable HTTP
Team server with /health, dynamic Server Card
Docker
Multi-stage build with health checks
Docker Compose
Production-ready with restart policies
Kubernetes
Manifests with Deployment, Service, Ingress
Helm
Chart with configurable values

Ready when you are

345 tools. 7 vendors. Production safety by default. Launching soon.

Coming Soon

Be the first to know

Follow @latticio on X for launch announcements. PyPI, Docker, and full documentation coming soon.