Customizing your project
Shape what your agents know and can do — agents, skills, connectors, secrets, channels, models, members, and settings.
Two places in a project shape how it works:
- Customize, in the project sidebar, opens the project's Connectors, Agents, and Skills pages.
- Settings, at the bottom of the sidebar, opens an overlay with a tab per area: secrets, channels, members, models, and more.
Agents, skills, triggers, and connector definitions are saved to your project
as code (in kortix.yaml and the agent config under .kortix/opencode/), so
they're versioned and travel with the repo. Secret values and credentials are
never written to the repo.
Agents
Open Customize → Agents to browse the project's agents and read each one's instructions. Agents are the personas your sessions run as — shaped around how your team actually works. Every project ships with sensible defaults so it works out of the box; refine them as you learn what you want. See Agents.
Skills
Open Customize → Skills to see the skills available to your agents. A skill packages a workflow you repeat into a reusable shortcut the agent can invoke. Open one to read its documentation and what it does.
Connectors
Open Customize → Connectors to connect the external tools your agent should act in — Gmail, Notion, GitHub, and many more — so it can do real work without you in the loop. Every agent in the project can use every connector you connect. For each connector you can:
- Connect credentials (a one-click connect for supported apps, or your own token).
- Set permissions per action — always run, require approval, or block — with rules that match specific tool calls.
- Set a project-wide default for tool calls no rule matches, then carve out exceptions.
See Connectors for setup details.
Secrets
Open Settings → Secrets to manage the API keys, tokens, and config your agents need. Add a value and choose its delivery — where the value is allowed to go. A Sandbox secret reaches every agent as an environment variable; stricter modes keep the value outside the sandbox. Secrets are encrypted at rest and never written into the manifest.
See Secrets for the full walkthrough.
Channels
Open Settings → Channels to drive the project from where your team already talks, so anyone can kick off work with a message. You can connect Slack, Inkbox (iMessage, text, email, and calls through your own Inkbox account), and Glyph live voice. Microsoft Teams appears there too once you turn it on under Settings → Experimental. See Slack & channels, Inkbox, and Glyph.
Models
Open Settings → Models to connect provider keys or your own ChatGPT, Claude, or Cursor plan, choose which models the picker offers, and see request logs. See Models.
Members
Open Settings → Members to control who can access the project and what they can do. Each person has a project role — Member, Editor, or Manager. See Accounts & tokens.
Other settings
- General — the project name and the sandbox provider sessions run on.
- Repositories — where the project's code lives, and how to work on it from your computer.
- Schedules and Webhooks — the project's triggers.
- Sandbox templates and Snapshots — the machine a session runs on. See Runtime.
- Experimental — features you can switch on before they are generally available.
Because agents, skills, and triggers are saved as code, you can review a change like any other, roll it back, and see exactly how a project was configured over time.
Related: Secrets · Connectors · Agents · Accounts & tokens.