Set up your AI
Toby uses an LLM for chat, summarization, and organization. You need at least one AI provider configured before chatting.
Supported providers:
- OpenAI (direct) — API key from platform.openai.com; models like
gpt-5-mini - Vercel AI Gateway — one Vercel key; models like
openai/gpt-5-minioranthropic/claude-sonnet-4.6
See the AI providers overview for how to choose between them and which models Toby recommends.
Open the configure UI
toby config
(toby configure is an alias for the same command.)
Under AI, add credentials for at least one provider (steps in the provider pages above). Toby stores keys in ~/.toby/credentials.json, not in your shell history.
Default providers (optional)
In the same configure UI, open Default Providers to pick which connected integration Toby prefers for each category when you do not specify one:
| Category | Example integrations |
|---|---|
| Calendar | Apple Calendar |
| Tasks | Todoist |
| Chat | Slack |
| Contacts | Azure AD |
| Search | Web Search |
| Work Tracker | Jira |
These defaults help schedules and multi-integration chat pick the right tools.
Personas and models
Each persona can use its own AI provider and model. The built-in Toby persona defaults to OpenAI (openai / gpt-5-mini). You can change provider and model per persona under Personas in configure.
For recommended models per provider, see OpenAI (direct) and Vercel AI Gateway.
Pretreatment follows the active persona’s provider (or your default persona when no session persona applies). Default pretreatment models are gpt-4.1-nano (OpenAI) or openai/gpt-4.1-nano (gateway). Override with TOBY_PRETREAT_MODEL. Planning uses its own model (gpt-4.1-mini). Set TOBY_PRETREAT_DELTA=0 to disable follow-up delta pretreatment.
Back up your configuration
Before moving to a new machine, create an encrypted backup:
toby config backup
toby config backup ./backups
Restore with:
toby config restore ./backups/your-backup.tbybak