Visual diagrams showing how Meta Council's core systems work -- outreach pipelines, council sessions, workflow DAGs, and scheduled automation.
Automated outreach sends personalized emails daily, then classifies inbound replies with AI to trigger follow-up actions, Slack alerts, and status updates.
Emails are sent daily at 9am PT. Inbound replies are parsed by SendGrid, classified by AI, and routed through configurable trigger rules.
A user query fans out to N expert agents running in parallel. Each agent produces an independent analysis, then a synthesizer merges all outputs into a single decision document.
Agents run in parallel with independent tool access. The synthesizer combines weighted outputs into a structured decision document with a permanent shareable URL.
Users describe a goal in natural language. The system parses intent, matches a workflow, confirms with the user, then executes steps sequentially with optional human-in-the-loop checkpoints.
Workflows execute as sequential DAGs. Checkpoint steps pause for human review before continuing. Each step's output feeds into the next agent's context.
Users attach cron schedules to workflows. A scheduler checks every 60 seconds for due runs, executes the full pipeline automatically, emails results, and computes the next run time.
The scheduler runs as a background task, checking every 60 seconds. When a schedule is due, it auto-executes the entire workflow pipeline and emails the results.