Agentic operating system · Open beta

A harness and database for always-on agents.

Orqa OS runs your agents. Orqa DB is the company repository. Together they are the Agentic Operating System.

How it works

Orqa OSOrqa DBOpen beta

How it works

Orqa OS runs the agents. Orqa DB is the company data lake.

Orqa OS is the harness. Orqa DB is where the company's data lives, and where the agents write to and read from.

Orqa OS · HarnessOpen beta

Orqa OS

A safer agentic harness for autonomous agents, built for teams who know their use case deeply and want to ship it without rebuilding the orchestration layer.

Orchestrator agents that are always on: running tasks, completing jobs, chasing goals 24/7.
Safety-first controls around what an agent may do.
Model-agnostic, tested across AI labs so the trial-and-error is already done.

Orqa DB · Company knowledge repoOpen beta

Orqa DB

A repository for everything a company knows, built for agent accessibility and visibility.

Document repository, data links and agentic interaction captured, indexed and graphed.
Designed like a company-pedia or library. Accessible via CLI, HTTP API and MCP.
Installs like a database: one settings file, orqa start, a port.

Orqa OS · Harness · Open beta

Orchestrator agents that run 24/7, with better visibility and security.

A safer agentic harness for autonomous agents, built for teams who know their use case deeply and want to ship it without rebuilding the orchestration layer.

Orchestrator agents, unlike chat models, are always on: running tasks, completing jobs, chasing goals 24/7.

We use the harness internally to test model suitability across AI labs, so clients don't have to do the trial-and-error themselves.

Tested across

Orqa OS · console

Orqa DB · Company knowledge repository · Open beta

An agentic Knowledge Database.

Orqa DB unifies a company's distributed data into a single knowledge graph with stable node identity and indexed edges. References, tags, block links and typed attributes are resolved into edges at write time, so backlinks, attributes and graph structure are always current. Every data element carries properties that indicate location, reference and lineage.

Books in a libraryKnowledge is shelved as books: a primary per organisation and any number of secondaries for corpora, reference material or mounted external systems. Books are isolated by default and independently addressable, joined only where an agent writes an edge.

Nodes and edgesPages, blocks and media are first-class nodes with stable identifiers. References, tags, block links and attributes written in ordinary text are resolved into edges on save, so backlinks, attributes and graph structure are current at all times.

Made for agentsBuilt-in reports surface the shape of the graph: attributes in use, value drift, undescribed entities, orphan pages and clusters. Two bundled skills teach an agent the syntax and the working patterns.

Files are nodes tooPDFs and links are ingested as pages with backlinks, attributes and notes. The same file or the same URL always returns the same node. Large files are supported, and archived ones go cold.

Nothing vanishesDeletes are soft and restorable for the retention window you set. Archiving is an owner-run sweep, and Orqa DB never removes or archives anything on its own.

Open at the edgesLossless export and import of a whole book. Markdown and JSON for documents, GraphML and DOT for Gephi, Graphviz and other graph tools, or a data lake.

Structured dataAn MCP connector to Postgres, so agents reach structured records through the same interface as the graph.

orqa · quick start

$ orqa start                 # the server, in the background
$ orqa status                # running? which books, which tokens
$ orqa page create --title "Alpha"
$ orqa block append --content "[[Alpha]] saved for later"
$ orqa media add-link --url https://example.org
$ orqa query search --q "alpha" --since 7d
$ orqa report attrs --since 30d  # what the graph looks like
$ orqa mcp --scope write         # connect an agent host

Installs

Like a database

Deploys from a release archive with one settings file and one command, orqa start. Nothing is compiled on your machine.

Runs on

Your machine or a container

Windows, macOS, Linux and Docker. Registers with the OS on request and returns after login or reboot.

Reached by

CLI · HTTP API · MCP

Named access tokens, each scoped to the books it may touch. A single MCP entry connects any compatible agent host.

Content model

Block based, no fixed schema

Block based, the way Notion's is. Attributes are written into the content as you go, so there is no schema to define first.

Works with

Works with Fidelity Agentic.

Fidelity Agentic

ISO/IEC 42001Trace engineLive

The compliance, risk and control layer for agentic ecosystems. One trace engine over every agent you run: every response, tool call and inter-agent exchange recorded, with policy applied at the point of action.

fidelityagentic.com

Fidelity Agentic trace, live — **Trace · live**

Fidelity Agentic policy gate — **Policy gate**

Status and access · Open beta

Orqa OS and Orqa DB are in open beta. Request access and you get:

01The release archive for Windows, macOS, Linux and Docker. Nothing is compiled on your machine.
02The documentation set: CLI guide, install guide, and the skills an agent reads.
03A seat in the Discord for questions and discussion, with the team that builds it.

A person reads every request · Mon–Fri, 09:00–17:00 AEST