Install

Up and running in five minutes.

One command, your infrastructure, your models — then see exactly what it looks like running. Every tab below is a real deploy. Click any image to expand.

curl -sSL https://install.openagentics.io | bash

One command, a guided wizard

The install wizard — Docker Compose path, AWS Bedrock provider.

Install wizard: choose Docker Compose or Kubernetes — Deploy target.

Install wizard: AWS Bedrock IAM credentials, secret masked — Provider creds, masked.

Install wizard: choose which tools to enable — Tools.

Install wizard: the stack is up and the wizard hands you the URL — Up — it hands you the URL.

The OpenAgentic sign-in screen on the running deploy — Sign in.

Docker Compose — the whole stack on one box

The running stack — docker compose ps, health, logs, then an admin API key creating a flow. A real session.

The full openagentic Docker Compose stack running, every service healthy — The whole stack, up.

OpenAgentic chat answering via AWS Bedrock, running locally in Compose — Chat, on your own keys.

A tool executing inside the local Compose stack — A real tool executing.

Kubernetes — your cluster, live

kubectl get pods — 12/12 Running — ingress, logs, then an admin API key creating a flow and a Bedrock chat session. A real session on the cluster.

The Kubernetes Lens dashboard showing every OpenAgentic pod Running in the openagentic namespace on the live cluster — Every pod `Running` in the `openagentic` namespace.

OpenAgentic chat dispatching the Prometheus and Kubernetes tools in parallel against the live cluster — Two tools, one prompt — the agent operating its own cluster.

OpenAgentic chat answering on AWS Bedrock Claude Sonnet 4.6 with an extended-thinking block — Chat on AWS Bedrock, embeddings on in-cluster Ollama.

Admin console — operate the platform itself

One console governs the whole platform — providers, models, routing, agents, tools, identity, and the audit chain. Every panel writes to the database, not a config file you redeploy.

Providers & models

Register Ollama · Bedrock · Vertex · Azure · OpenAI; assign a model per capability (chat, embed, image) with capability flags. llm_providers

Routing & tuning

Set the SmartModelRouter thresholds, the chat-loop turn ceiling, token budgets, and compaction points — live. system_configuration

Agents & tools

The agent roster (role · provider · model) and the full tool fleet, each with its capabilities — enable, scope, inspect.

RBAC & permissions

Roles and per-role permissions decide who can run what, gating every tool call by the local account's role. (Per-user SSO/OBO identity is an enterprise feature.) rbac

Audit chain

Every tool call lands in an append-only, tamper-evident log — who, what, when, and the result. tool_call_audit_log

Built-in assistant

An in-console assistant that advises on model cost/quality tuning and explains what each setting changes.

The admin console tool dashboard: every bundled tool with its capabilities — The tool fleet — every bundled tool, with its capabilities.

The admin console Agent Registry: the full roster of agents with their roles, providers, and models — Agent registry.

The admin console Models manager editing a Claude Sonnet 4.6 model on AWS Bedrock with capability toggles — Model registry + capability flags.

The admin console with the built-in assistant advising on model cost tuning — Built-in tuning assistant.

Cloud & tools — it operates your stack

OpenAgentics agent running three tools in parallel for a cross-domain review — Kubernetes + AWS + web, fired in parallel.

The agent's Kubernetes findings: a real table of deployments and pods all Running — Real cluster state, in a table.

The agent's synthesized report: real AWS spend by service and live pricing with cited web sources — Cited cost report — real spend + live prices.

OpenAgentic chat pulling real AWS spend via the AWS tool and rendering a Sankey cost-flow diagram — Real Cost Explorer data → live Sankey + KPI dashboard.

OpenAgentics human-in-the-loop approval modal pausing a mutating tool call — Human-in-the-loop — mutating calls wait for Approve / Deny.

Flows — a no-code builder, extensible

OpenAgentics Flows: the visual workflow builder with a palette of agents and ready-made templates — The Flows builder — agents, templates, drag and wire.

A flow executing — the web-read node completed, the Structured Brief node running — Running.

The flow completed — every node green, the structured brief streamed into the output panel — Completed — grounded output.

The flow's final structured-brief JSON output with TL;DR, key points, entities and open questions — The structured brief, cited.

LangGraph
LangChain
n8n import / export
Flowise compatible
Signed run traces
Human-in-the-loop gates

Code — your repos, your IDE, your way

No plugin to install, no editor to marry. OpenAgentic reaches into your real repositories through the GitHub tool — under your own token and projects — so it works in and with the code you already have, while you keep whatever IDE you already use.

A built-in code-execution sub-agent runs generated scripts in a sandbox to compute, transform, and verify inline — behind the same approval gate and audit log as any other action. See Code & repos in the docs. (The desktop Gnomus Code apps and Gnomus Synth on-the-fly tool synthesis are AgenticWork products, not part of the open-source edition.)