The agent workflow
- 1
Give the agent its own member
Invite the agent as its own Tiqzen member and add it only to the project it owns. It can see and change exactly what that member can — enforced server-side, not by prompt.
- 2
Connect the agent over MCP
Add the Tiqzen MCP server URL to Claude Code, Cursor or any MCP-compatible agent, then sign in once as the agent member.
- 3
Let the agent pull its queue
The agent calls list_tickets to find work — filtered by status and severity — instead of being spoon-fed tasks.
- 4
Close the loop automatically
As the agent fixes issues it calls update_ticket to move status, reassign, or lower severity — the backlog stays truthful.
Agent updates land on the board humans watch
An autonomous agent with workspace-wide write access is a liability — so an agent acts with the permissions of the member it signs in as, enforced at the API layer. Give it a member that belongs to one project and it works that lane, moves its tickets, and the team sees it happen in real time.
Backlog · 4
In progress · 2
Done · 7
Agents pull their queue
Agents pull work by severity instead of waiting for assignment.
Auditable updates
Status changes are auditable — every update is attributed to the agent’s member in the audit log.
One shared board
Humans keep the same board: agent updates appear in real time alongside the team's.
See pricing → · Back to the MCP server →
Hand an agent its first project
Full MCP access ships in the Free plan — try an agent on a single project before rolling it out to the team.