AI builds the workflow. A deterministic engine runs it.

---

AI helps you build. It never runs in the execution path, so the same workflow behaves the same way every time.

Every version must pass simulation and human approval before it can run against production devices.

SSH, NETCONF, SNMPv3, REST, Ansible, Python and GitOps in one workflow.

Runs on your infrastructure. Connect OpenAI, Anthropic, Gemini, DeepSeek, Kimi or a local model through Ollama.

Tell the assistant what you need in plain English, such as "list active devices from NetBox and email a CSV report."

The assistant drafts the workflow on the canvas: steps, targets, and success and failure paths. Edit any node by hand.

A simulation validates the graph and is bound to that exact version. Change a node and it has to pass again.

Promote from Draft to QA to Production. Production requires the right role and a human approval.

The engine executes step by step with live status, per-step logs, retries and a complete audit trail.

Follow every run on a live graph. Open any step to see its logs and its input and output payloads.

Schedules and triggers: run on a timer, on demand or from a webhook.

Reports: CSV and HTML output delivered by email or chat.

Administrators set the rules once and FlowWeaver enforces them on every workflow, run and AI action.

Role-based access for building, approving and running.

Global policies as guardrails across all workflows.

Audit log of every change and every assistant action.

Cisco IOS XE, Juniper Junos, Arista EOS, Nokia SR Linux and any device reachable by SSH, NETCONF, SNMPv3 or REST.

NetBox, Infoblox and ServiceNow, plus any REST API through integrations.

Git repositories, webhooks, and workflow import and export.

Slack, Microsoft Teams, Telegram, WhatsApp and email.

OpenAI, Anthropic, Google Gemini, DeepSeek, Kimi, Ollama (local) or a custom endpoint.

MCP servers and reusable skills extend what the assistant can do. Third-party names are used for identification only.