Each Custom Agent has its own independent set of enabled tools and connected apps. This lets you build focused agents with exactly the capabilities they need — and nothing they don’t.Documentation Index
Fetch the complete documentation index at: https://docs.zerotwo.ai/llms.txt
Use this file to discover all available pages before exploring further.
Why Per-Agent Configuration Matters
Restricting tools and connectors per agent keeps agents focused and reduces the risk of unintended actions:- A coding assistant doesn’t need image generation or calendar access
- A research bot doesn’t need database write access or Slack
- A writing helper doesn’t need GitHub or Supabase
Configuring Tools
In the agent editor (/agents → select agent → Edit), toggle tools on or off:
| Tool | What it does |
|---|---|
| Web Search | Agent can search the internet for current information |
| Deep Research | Agent runs extended multi-source research sessions |
| Canvas Mode | Agent can create structured documents in the canvas editor |
| Agent Mode (autonomous actions) | Agent can take autonomous multi-step actions (like Agent Mode in chat) |
| File Uploads | Users can attach files to chats with this agent |
| Voice | Agent can respond with voice output |
Configuring Connectors
In the agent editor, select which of your connected OAuth apps this agent can use. The dropdown or checkbox list shows only apps you’ve already connected in Settings → Connectors. Example configurations:| Agent purpose | Suggested connectors |
|---|---|
| Daily planner | Google Calendar, Gmail |
| Dev workflow helper | GitHub, Linear, Notion |
| Content manager | Google Drive, Notion |
| Data analyst | Supabase, Google Sheets |
| Email manager | Gmail |
| Research assistant | None needed (uses web search) |
Permission Model
The agent uses your OAuth credentials for every connected app. It inherits your actual permissions from each connected app — it cannot do more than you’re authorized to do. If you connected GitHub with read-only access, an agent using that GitHub connector can only read — it cannot create issues or open PRs, even if you try to prompt it to do so.Multiple Agents with Different Scopes
You can create multiple agents, each with a different set of connectors, to handle different workflows securely:Best Practices
- Start minimal — enable only the tools and connectors the agent genuinely needs. Add more as you find specific use cases that require them.
- Test after configuring — start a chat with the agent and confirm it can access what it needs (and is blocked from what it shouldn’t have).
- Re-authorize expired connectors — if an agent reports it can’t access a connected app, the OAuth token may have expired. Go to Settings → Connectors, disconnect and reconnect the app, then return to the agent settings and reselect it.

