Agentic FabriqCoding Agents Administration
Enable, configure, and monitor coding agents for your organization.
Overview
Organization admins control which coding agents are available to members. Each agent must be explicitly enabled before any member can use it, and active broker sessions are visible and can be revoked at any time.
- Admins control which coding agents are available to organization members
- Each agent must be explicitly enabled before members can use it
- Active broker sessions are visible and can be revoked
Enabling a Coding Agent
Follow these steps to enable a coding agent for your organization:
- Navigate to Admin Console > Coding Agents
- Click "Enable" on the agent you want to activate (Claude Code, Cursor, or Codex)
- The agent appears in the enabled list with a "Pending" status until a member uses it
- Copy the MCP config snippet and share it with your team
Tip: The MCP config snippet can be copied directly from the agent's detail page. Share it with team members so they can configure their local environment.
Supported Agents
The following coding agents are supported:
| Agent | Description | MCP Config |
|---|---|---|
| Claude Code | Anthropic's Claude-powered coding agent | {"mcpServers": {"agentic-fabriq": {"command": "afctl", "args": ["broker"]}}} |
| Cursor | AI-first code editor with built-in agent | Same as above (or full path for Remote SSH) |
| Codex | OpenAI's coding agent | Same as above |
Managing Sessions
Admins can view and manage active broker sessions for each coding agent.
Viewing Active Sessions
Navigate to Admin Console > Coding Agents, click on an agent, then select "Manage Sessions". Each session displays:
- Session ID — Unique identifier for the broker session
- User — The organization member who initiated the session
- Created — Timestamp when the session was established
Revoking a Session
Click "Revoke" on any active session to immediately disconnect it. The broker on the user's machine will need to re-authenticate.
Note: Sessions auto-expire approximately 3 minutes after the broker process exits (heartbeat-based TTL). You only need to manually revoke sessions if you want to force an immediate disconnect.
API Reference
The following REST endpoints are available for managing coding agents programmatically:
| Endpoint | Description |
|---|---|
GET /api/v1/coding-agents | List all supported agents with enabled/disabled status |
POST /api/v1/coding-agents | Enable a coding agent (body: {"app_name": "claude-code"}) |
DELETE /api/v1/coding-agents/{agent_id} | Disable a coding agent |
GET /api/v1/broker-sessions | List active broker sessions |
DELETE /api/v1/broker-sessions/{session_id} | Revoke a session |
Example: Enable a Coding Agent
curl -X POST https://dashboard.agenticfabriq.com/api/v1/coding-agents \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{"app_name": "claude-code"}'Scope Management
After enabling a coding agent, configure which tool scopes it can access. Scopes define the boundaries of what the agent is permitted to do on behalf of users.
- Navigate to the agent's detail page and click "Manage Scopes"
- Assign scopes matching the tools your team uses (e.g.,
gmail.send,slack:chat:write) - The agent can only access tools and scopes explicitly granted by the admin
Tip: Start with a minimal set of scopes and expand as needed. You can always add more scopes later without disrupting active sessions.
Security Considerations
- Each coding agent operates with the permissions of the individual user who authenticated
- Admin-assigned scopes act as an upper bound — a user cannot access more than what the admin allows and what the user has personally connected
- All tool invocations are logged in the audit trail
- Revoking a session immediately disconnects the broker
Important: Even if an admin grants broad scopes, a user can only use tools they have personally authenticated with. The effective permission set is the intersection of admin-assigned scopes and user-connected services.
Need help?
Our team is here to help you get started.