Orchestri runs on infrastructure you control. Here’s what you need to turn the demo into your own working setup.
You’ll need access to the Orchestri source repository. This website doesn’t create an account or provision an installation.
A public source download is not linked here yet. If you’ve been given repository access, use the installation guide included in your checkout.
The reference installation uses Linux, PostgreSQL, and either systemd services or Docker Compose. Node.js builds the web interface; Python runs the core services.
Read docs/INSTALL.md in the source checkout for the database, secrets, authentication, and service setup. Start by inspecting the helper:
bin/orchestri-setup --helpCompose starts the application services. Agent runtimes are a separate setup step.
Codex: set up the reviewed Linux aarch64 binaries and AppArmor policy, then sign in with ChatGPT through Orchestri. See docs/codex-agent-setup.md.
Hermes: connect an installed Hermes runtime and configure managed agents. See docs/MANAGED_AGENTS.md.
Deep Agents: prepare Docker and the sandbox image. See docs/deep-agents-runtime.md.
Configure a provider, verify a model, and publish an agent. Create a task, describe the outcome, and choose the agent that should work on it.
Select a registered repository for project changes, or use a Codex scratch workspace for a general task. Start the task, answer questions, and review what comes back.
Gmail, Slack, Notion, Jira, and recurring automations can come later. Your task store works without them.
This marketing site and your application are separate deployments. Keep your app’s authentication and access controls in place when linking to it.
Explore the example workspace