How to use Odysseus after installing it

1. Log in and confirm the workspace loads

Use the admin credentials from your first-run logs. If the login screen rejects them, stop and fix credentials before configuring models or tools.

Fix invalid credentials

2. Connect a model provider

Odysseus needs a model backend before chat, agents, and research feel useful. For local beginners, Ollama is usually the fastest path.

ollama pull qwen2.5:7b
ollama list

If Odysseus runs in Docker while Ollama runs on the host, use the host gateway URL:

http://host.docker.internal:11434/v1

3. Test a simple chat first

Before adding agents or MCP tools, send one simple prompt and confirm the selected model responds. This isolates model connectivity from higher-level workspace issues.

4. Use agents only after chat works

Agents make multiple model and tool calls. If the base model is slow, missing, or too small, agents will feel broken. Use a stronger model for planning and multi-step work when your hardware allows it.

5. Add one MCP tool at a time

MCP can connect Odysseus to files, GitHub, databases, cloud drives, calendars, and more. Start with a narrow tool and limited permissions so you can understand what the agent can access.

• Use read-only permissions first when possible.
• Restrict filesystem roots to test folders.
• Never paste broad production credentials into a new tool test.
• Review logs after each tool call.

6. Use Deep Research with search configured

Deep Research needs reliable search. If SearXNG returns empty or blocked results, fix search before judging Odysseus research quality.

Set up SearXNG for Deep Research

A safe first-day routine

1. Open Odysseus locally.
2. Log in and save credentials securely.
3. Connect one model provider.
4. Test basic chat.
5. Add one MCP tool with limited scope.
6. Try one short research workflow.
7. Back up your config before exposing anything outside your LAN.