This guide covers practical deployment shapes for Orchestro beyond “run it once in a shell.”
- Supported Deployment Mindset
- Local Persistent Service
- Data Placement
- Backend Dependencies
- Recommended Validation After Deploy
The codebase is currently best suited to:
- local personal deployment
- a private workstation or homelab box
- an operator who can inspect logs, files, and SQLite state directly
It is not yet documented or hardened as an internet-facing production SaaS deployment.
The simplest long-running deployment is:
export ORCHESTRO_HOME=/path/to/orchestro-home
orchestro serve --host 127.0.0.1 --port 8765Concrete examples now live under:
Persist:
- the Orchestro home directory
- the project repo if you rely on project-local instructions or constitutions
Important files under ORCHESTRO_HOME may include:
orchestro.dbglobal.mdconstitutions/mcp_servers.jsonlsp_servers.jsontool_approvals.jsontrust.jsonplugins/
A deployment is only as useful as its configured backends.
Common dependency classes:
- local OpenAI-compatible server
- agent CLIs present on
PATH - Anthropic/OpenAI/OpenRouter API credentials
- MCP server subprocess dependencies
- language server binaries
Use these to inspect the deployed environment:
orchestro backends
orchestro plugins
orchestro mcp-status
orchestro lsp-statusMinimum checks:
PYTHONPATH=src pytest -q
PYTHONPATH=src python scripts/orchestro_canary.py --backend mock --json
curl http://127.0.0.1:8765/health
curl http://127.0.0.1:8765/backendsIf you rely on live backends:
ORCHESTRO_RUN_LIVE_BACKEND_TESTS=1 PYTHONPATH=src pytest -q tests/test_live_backends.pyRelated docs: