What orqo is, who it is for, and the things people ask before they start. Each answer stands on its own — open the ones you need.
orqo is a system, in the cloud, where teams of humans and their AI agents work together as one organization. You describe a complex job once in plain language; orqo builds the multi-agent workflow to run it, then runs it again and again, getting sharper each pass.
Most companies manage AI as a patchwork of scattered services and subscriptions, one per employee, each with its own memory and nothing anyone else can reuse or check. orqo is one system the whole organization runs. You describe the work in plain language, orqo builds the multi-agent workflow that runs it, and a colleague who has never thought about AI runs it from a single button. Underneath that: a typed knowledge graph built from your own documents (57 knowledge types, 48 typed relations from educational-science research) that agents navigate by meaning rather than search, and that you can publish as a chatbot or an MCP server; integrations orqo writes on demand for systems no catalog carries; workflows that mix agents where you want judgment with deterministic logic where you want certainty; per-agent context compaction instead of one catastrophic summary; and split-plane architecture where the control plane holds none of your content and execution can run in your own cloud or on hardware we deliver.
Both ends of the spectrum — today. A solo operator can run like a whole team; an enterprise can put hundreds of people and their agents on one governed system, with its own knowledge ingested into a private graph and workflows that repeat and sharpen with every run. The line isn't individual-vs-team — it's recurring, governed work versus one-off prompts. Individuals self-serve in minutes; for organizations we also offer guided onboarding — a hands-on setup to ingest your material and stand up your first workflows under a B2B agreement. One system underneath, whether you're one person or ten thousand.
orqo is in sneak preview — a free, limited early-access phase. A small number of slots are open so that the people and teams putting orqo to real work help shape it before general release. During the preview you bring your own LLM keys and pay your provider directly; paid subscriptions, with usage credits included, switch on when the preview graduates.
Two different things share that question. LangChain is a code library you wire together by hand. CrewAI has grown past that into a real platform — hosted service, visual builder, enterprise tier — so calling it 'just a framework' undersells it, and on the build-the-agents axis it and orqo genuinely overlap (teams of agents, orchestrated workflows, any model per agent, bring-your-own inference). The difference is the buyer and the unit. CrewAI is where a developer builds agents; orqo is where an organization runs them: many people and their agents co-working in one governed system — roles, sharing, and audit across the company — on a typed knowledge graph grounded in your own content (embeddings get agents in the door; typed relations then let them navigate by meaning, not nearest-neighbor alone — and never from an AI-written summary of your sources), with engineered context management, real two-way integrations it builds on demand, and a Chief of Staff reachable on Slack, WhatsApp, and email. You drive it by describing the work in plain language — and when you do want code, a granted member gets a full in-browser Python environment, so it's no Python required, not no Python at all.
With declared roles and named outcomes, rather than instructions buried in a prompt. A stage can carry a Finalizer: an assignment whose job is to judge the result the other agents produced. The stage then routes on the outcome the Finalizer returns — a 'fail' sends the work back into the same stage to be redone, a pass moves it on. That is the evaluator-optimizer pattern the field has converged on (something produces, something else judges, another pass until the verdict changes), except it is configured on a canvas instead of written into a prompt or a while-loop in someone's script. And because every pass is a run on the record, you can see how many iterations a result actually took and what each one cost, which is what makes a loop safe to put in production rather than a mystery line on the bill.
Yes. orqo is model-agnostic — you bring your own LLM provider keys, encrypted and injected at runtime, and pay your provider directly for model usage.
Yes — orqo is split-plane by design. The control plane is always orqo's SaaS and holds none of your content; the execution plane that processes your data runs in orqo's cloud by default, in your own cloud, or on hardware orqo delivers to your premises.
Preview is free. You bring your own LLM keys and pay your model provider directly for usage.
Preview is free. You bring your own LLM keys — encrypted and injected at runtime — and pay your provider directly.