CLI
One tool. Two ways to run it.
Oynix installs as a single command. Pick the build that fits: the local CLI runs entirely on your machine, and the server CLI runs a shared engine your whole team queries from. The commands you learn carry across both.
The local CLI
Runs entirely on your machine. One person, no account, no Docker or server.
- The engine auto-starts. You rarely type it.
- Your graph stays on your machine, at ~/.oynix/local.
- Start key-free with a structural graph. Add a model later in the Models UI and re-index.
- A lean menu: init, connect, graph, status, doctor.
Best for trying Oynix, solo work, or a private repo on your laptop.
install → oynix init → oynix connect → oynix graph → ask in your agentThe server CLI
A shared team engine. Roles, workspace keys, and a web console, on a Docker stack or your own server.
- Runs the Docker engine: graph, indexer, and ingestion together.
- Sign in, bind a workspace key, add members with roles.
- One shared memory the whole team queries from their agents.
- Ask from your agent or from Slack, plus architecture checks as a CI quality gate.
- Governance, keys, and members live in the web console.
Best for a team that wants one shared memory, access control, and CI.
| Goal | Local | Server |
|---|---|---|
| First-time setup | init → connect → mcp setup | Sign in, create the workspace, start the shared engine |
| Add code and docs | The connectors page (oynix connect) | The connectors page, shared across the workspace |
| Set a provider key | Models UI | One workspace key, held once for the whole team |
| Ask about your code | In your agent, over MCP | In your agent, in Slack, or in the console |
| See the graph | oynix graph | The graph view, on the shared engine |
| Stay current | engine update | Updates in place, data preserved |
| Quality gate in CI | not in local | Architecture rules enforced on every pull request |
| Team keys and members | not in local | Roles and members in the web console |
| The wiki | not in local | A handbook written from the graph, with citations |
The menus stay small on purpose. Adding code lives on the connectors page, keys and roles live in the web console, and answers come from your agent. Every command, flag, and output is written up in the full reference.