Flows Examples

Memory & Obsidian

Built-in Persistent Memory

SwarmOS uses domain-based memory: agents in the same domain share context. Code agents share architecture patterns; content agents share campaign playbooks; support retains customer resolution history. Agents read memory at task start and write new learnings at task end—memory persists across sessions.

Obsidian Integration

Connect an Obsidian vault as your knowledge base. At task start, agents read relevant notes; at task end, they update notes with new learnings and patterns. Configure per-org (Settings > Knowledge Base) or per-agent (Agent Manager > Edit Agent > Knowledge Base). Use [[wikilinks]] between notes. Agent config overrides org default.

Marketing Flow

Agents: marketer (coordinator), researcher, social

1. Create task: "Run Q3 campaign: competitor analysis, blog post, social threads"
2. Coordinator breaks into subtasks: researcher does competitive intel; marketer drafts blog; social creates threads
3. Memory: researcher writes to Research notes; marketer reads brand playbook from memory
4. Obsidian: campaign notes in Marketing/Campaigns; learnings linked with wikilinks
5. Review: approve drafts; "Implement Fixes" refines copy with feedback

Outcome: Multi-channel campaign with shared context. Next campaign reuses patterns from Obsidian.

Bug Finding Flow

Agents: jarvis (coordinator), coder-1/2/3, Senior Reviewer

1. Create task: "Investigate payment failures in checkout"
2. Coordinator routes to coder with debugging skill
3. Coder reads codebase, runs tests, checks logs; writes findings to agent log
4. Memory: code domain recalls similar past bugs (e.g., auth token expiry)
5. Obsidian: update Debugging/Payment-Checkout-Issues.md with root cause
6. Review: Senior Reviewer approves or requests changes; "Implement Fixes" re-runs if needed

Outcome: Faster diagnosis using shared memory. Future similar bugs surface past notes.

Customer Support Flow

Agents: support

1. Create task: "Customer X reports login issues after password reset"
2. Support reads customer history from memory and past tickets
3. Memory: support domain retains resolution patterns (e.g., cache clear + re-auth for session bugs)
4. Obsidian: Support/ FAQ notes, escalation runbooks, common fixes
5. Response: agent drafts reply using context; human approves before sending if configured
6. Learnings: update Support/Login-Issues.md if new pattern discovered

Outcome: Consistent, context-aware responses. Knowledge base grows with every ticket.

Content & SEO Flow

Agents: marketer, researcher, social

Researcher does keyword research; marketer drafts SEO-optimized article; social repurposes into threads. Memory shares SEO keywords and content calendar. Obsidian: articles in Content/Published/, templates in Content/Templates/.

DevOps & Deployment Flow

Agents: coder, jarvis (DevOps Engineer if configured)

Coder runs deploy script; quality gates (type, test, lint, build) auto-run. Deploy may require human approval. Memory stores deploy runbooks. Obsidian: DevOps/Deploy-Runbook.md, DevOps/Incident-Patterns.md.

Research & Competitive Intel Flow

Agents: researcher

Weekly competitor product update summary. Researcher scrapes/crawls competitor sites and release notes. Obsidian: Research/Competitor-Product-Updates.md—append weekly; [[Competitor A]] links to profile notes. Output: summary note + optional alert to marketer/social.