Jacomeit IT
Since July 2026
AI agent team in daily operations · Own initiative
Seven agent roles connected to Jira, Confluence and Slack via MCP, plus 34 active n8n workflows
Context: An own initiative, not a client engagement. Since July 2026, research, drafting, operations monitoring and reporting in my own business run through a team of specialised AI agents. Each role has its own profile, its own Atlassian account and its own Slack channel.
My Role
Design, build and operation: role boundaries, tool access via MCP, workflow development in n8n, running the LLM gateway on my own server, and the approval and documentation rules the agents work under.
Highlights
- Seven agent roles, each with its own profile, Atlassian account and Slack channel
- 34 active n8n workflows: 11 Slack agents, 12 scheduled jobs, 10 sub-workflows, 1 chat
- 615 workflow executions between 1 and 15 August 2026, 7 of them failed — 98.9 % success rate
- Approval workflow as a hard rule: research and drafts autonomous, publishing and sending only after sign-off
- Architecture decisions recorded as ADRs in a decision log, not just implemented
- Own LLM gateway (LiteLLM) on own infrastructure, model chosen per role rather than one for all
Stack
Claude (Opus/Sonnet) · MCP · n8n · LiteLLM · Slack · Jira · Confluence · Baserow · Docker