v0.1.0 — open source

OculOS

Build AI workflows
visually

Drag, drop, and wire AI nodes into automated workflows.
Triggers, AI steps, tools, and logic — no code required.

$pip install oculos && oculos up

Get Started View on GitHub

Everything you need to build AI workflows

Trigger → AI → Tool → Logic. Wire it visually, run it instantly.

Build

Visual Workflow Builder

Drag and drop nodes onto a canvas. Wire triggers, AI steps, tools, and logic into a directed graph — no code required.

AI Nodes

Transform, Decide, Generate, Guard — four AI node types that let you run LLM steps anywhere in your workflow with a system prompt and model picker.

Connect

Tool Nodes

Built-in nodes for HTTP requests, email, database queries, file ops, and MCP tools. Connect to any external system without writing glue code.

Trigger

Triggers

Start workflows on a cron schedule, via webhook, or manually from the dashboard. One trigger node kicks off the entire graph.

Run

Live Execution Streaming

Every node lights up in real-time as the workflow runs. Server-sent events stream status — running, done, error — directly onto the canvas.

Observe

Run History & Output Viewer

Every execution is stored with full node-level output. Expand any run to inspect what each step produced, how long it took, and what it cost.

Secure

Secrets Vault

Store API keys your workflow nodes need — OpenAI, Anthropic, Slack, and more. AES-256 encrypted at rest, never logged.

Configure

Model Agnostic

Pick a different model for every AI node. GPT-4o for heavy reasoning, Haiku for fast classification — all in the same workflow.

Build

Branch & Loop Logic

Add conditional branches with Python expressions and loop nodes for iteration. Build real decision trees, not just linear pipelines.

Simple by design

Four integration paths. Pick the one that fits your stack.

Install & start

One command spins up the control plane and serves the dashboard at localhost:9090.

pip install oculos
oculos up

CLI commands

Inspect and manage agents without opening a browser.

oculos status          # server health + cost summary
oculos agents list     # all registered agents
oculos agents add      # manually register an agent
oculos agents remove   # deregister an agent

Running in 30 seconds

No Postgres. No org charts. No 400-line config files.

Terminal

# Install

pip install oculos

# Start

oculos up

# Open dashboard

# http://localhost:9090

SDK Middleware

One line for existing apps

from oculos_sdk.middleware import OculosMiddleware app.add_middleware(OculosMiddleware)

Standalone SDK

Any Python agent

from oculos_sdk import Oculos ax = Oculos(server="http://localhost:9090") ax.register(name="my-agent", model="gpt-4o") with ax.task("summarize", budget="$2.00") as t: t.report_cost(0.045, model="gpt-4o") t.complete(result)

Passive Mode

Zero code changes

# Register agents manually in dashboard # OculOS polls /health and /ws/trace

Why OculOS

Most workflow tools lock you into their cloud and their pricing.
OculOS runs locally, open source, no lock-in.

Feature	OculOS	Others
Visual node-based workflow builder
AI nodes (Transform, Decide, Generate, Guard)
Tool nodes (HTTP, Email, Database, MCP)
Schedule & webhook triggers
Live execution with real-time node streaming
Run history & full output viewer
Centralized secrets vault
Zero-config startup (SQLite)
Model agnostic — pick model per node
Open source, self-hosted

Build AI workflowsvisually