Agents
Chat with Claude Code, Codex, or Cursor inside Appenda. The Agents panel sits beside your table, keeps context on the table you have open, and runs Appenda table tools when the agent needs data or writes.
Open Agents from the title bar. The panel sits beside your table and defaults context to the table you have open.
Open the panel
Click Agents in the title bar (robot icon, left of Settings). The panel opens on the right while your table stays visible.
You can run several chats at once. Each tab is tied to one provider runtime.
What you can do
- Ask questions about the open table in plain language
- Let the agent call reads (
count_rows,table_window,search_cells, …) - Approve writes when your chat mode or approvals policy requires it
- Reopen past threads from the history menu (clock icon on the tab bar)
- Watch long jobs (list import, enrichment) from the status strip and stop them when needed
Pick a chat mode in the composer. When the agent calls a risky tool, an approval card may appear unless your mode and workspace policy already allow it.
Before your first chat
- Pick a provider when you start a new tab
- Complete sign-in when the adapter asks for it
- Confirm your role may use the mode you need (Permissions)
Agents require a signed-in paid or trial workspace at runtime.
Related
- Chat modes
- Approvals
- Permissions
- Providers
- Sign-in
- Connect with MCP for external agent wiring
- Tool catalog for the full
appenda.*list
FAQs
Is Agents the same as MCP?
No. Agents is the in-app chat panel. MCP connects an external Codex, Cursor, or Claude Code client to the same Appenda tool gateway over loopback.
Does the agent always see my current table?
New prompts default to the table tab you have open. The agent still discovers other tables through list_tables when your question spans the workspace.
Can I use Cursor ask/agent/edit modes here?
No. Appenda owns chat modes (Read only, Ask, Allow session, Full access). They are separate from Cursor's native modes.