Hermes Agent Automation Workflows: Cron, Webhooks, and Self-Learning Pipelines
TL;DR: Hermes Agent's built-in cron scheduler, webhook system, and self-learning pipeline let you automate everything from daily Slack summaries to CI/CD triggers — without writing glue code. This guide covers 3 automation patterns with real configs, safety rules, and a workflow that improves over time.
Last updated: May 12, 2026
Why Automate With Hermes?
Hermes Agent is designed to run continuously — that's what makes it different from interactive coding agents. Once deployed (see our production deployment guide), the agent doesn't sit idle between conversations. It can monitor your issue tracker, run scheduled research sweeps, respond to webhooks from GitHub or Stripe, and chain multi-step workflows that create and refine their own skills over time.Three automation primitives give you this power: cron for scheduled tasks, webhooks for event-driven triggers, and self-learning pipelines that improve with every run.
Pattern 1: Cron — Scheduled Tasks
Hermes has a built-in cron subsystem accessible through both CLI commands and interactive chat. You can define recurring jobs using natural language or standard cron expressions.
Adding a Cron Job via CLI
class="language-bash"># Daily research summary at 8 AM hermes cron add --schedule "0 8 * * *" \ --task "Scan Hacker News and GitHub trending for AI agent news. \ Summarize top 5 stories in a bullet list."Weekly performance report every Monday
hermes cron add —schedule “0 9 * * 1”
—task “Check my deployed agents’ memory usage and SQLite DB size.
Send a report to #ops channel.”
Adding via Chat
While chatting with Hermes on any messaging platform, use:
/cron add --schedule "0 6 * * *" --task "Check weather for San Francisco. \
If rain expected, remind me to bring an umbrella to the #general channel."Managing Cron Jobs
| Command | Description |
|---|---|
/cron list | Show all scheduled jobs with next run time |
/cron remove <id> | Delete a specific job |
/cron pause <id> | Pause without deleting |
/cron resume <id> | Resume a paused job |
/cron log <id> | View execution history for a job |
Safety Rule
Hermes enforces a critical guard: cron-initiated sessions cannot create new cron jobs. This prevents runaway scheduling loops where a job spawns infinite descendant jobs. If your daily research agent tries to schedule its own recurrence, Hermes blocks it.
Pattern 2: Webhooks — Event-Driven Triggers
Webhooks let external services ping Hermes when something happens. The agent runs an HTTP server that validates HMAC signatures, transforms payloads into prompts, and routes responses back.
GitHub PR Review Webhook
class="language-bash"># Configure webhook on your VPS
hermes webhook add \
--path /github-pr \
--secret "your-hmac-secret" \
--prompt "A new PR has been opened. Review the diff, check for \
security issues, and post a summary comment."https://your-agent:8443/github-pr) to your GitHub repository's webhook settings. When a PR opens, Hermes receives the payload, reviews the diff, and posts a review comment automatically.
Available Webhook Integrations
| Service | Trigger | Typical Agent Action |
|---|---|---|
| GitHub | PR opened, issue created, push | Review code, triage issues, run checks |
| GitLab | MR created, pipeline status | Review merge requests, analyze failures |
| JIRA | Ticket status change | Update context, start implementation |
| Stripe | Payment event, subscription change | Log transactions, flag anomalies |
| Linear | Issue created, status changed | Pick up work, auto-implement |
Pattern 3: Self-Learning Pipelines
This is Hermes's most distinctive automation feature. When the agent observes repeated multi-step workflows — daily research briefings, weekly PR triage, monthly dependency audits — it automatically creates a skill from the pattern. Over time, these skills become faster and more accurate because Hermes tracks which steps worked and which didn't.
How Self-Learning Works
- Repetition detection. Hermes tracks task types across sessions. When it sees the third "daily tech briefing" request, it flags the pattern.
- Skill generation. The agent writes a reusable skill file (a structured prompt + tool chain) that captures the workflow.
- Refinement. On each subsequent run, Hermes compares actual execution against the skill. Steps that consistently succeed become optimized; steps that fail trigger skill updates.
- Automation escalation. Once a skill reaches confidence threshold, Hermes offers to schedule it as a cron job — closing the loop from manual task to automated pipeline.
Example: Competitive Monitoring Briefing
Suppose you manually ask Hermes each morning: "What did Anthropic and Google announce yesterday?" After three days, Hermes creates a skill calledcompetitive_briefing. By day seven, the skill runs in under 2 minutes instead of 8. By day fourteen, Hermes offers to schedule it as 0 7 * * 1-5 — a fully automated weekday briefing.
class="language-bash"># After two weeks, this runs automatically at 7 AM weekdays
hermes cron list | grep competitive
# → ID: cron_12 | competitive_briefing | Mon-Fri 7:00 AM | 14 runs | 62s avgPutting It Together: A Complete Daily Workflow
Here's what a Hermes-powered engineering team's automation stack looks like:
| Time | Task | Pattern | Output |
|---|---|---|---|
| 6:00 AM | Scan GitHub trending + HN for AI news | Cron | Slack #research channel |
| 7:00 AM | Competitive briefing (Anthropic, Google, Meta) | Self-learning pipeline | Notion doc |
| 8:00 AM | Review overnight PRs | GitHub webhook | PR comments + Slack digest |
| 9:00 AM | Team standup summary from Slack messages | Cron | Email to team lead |
| Continuous | Triage new Linear tickets | Linear webhook | Agent picks up implementable tasks |
| 10:00 PM | DB health check + skill audit | Cron | Ops dashboard update |
FAQ
Can cron jobs send results to different platforms?
Yes. Each cron job can route output to a specific Slack channel, Discord channel, Telegram chat, email address, or local file. Configure the--output flag when adding the job.
Are webhooks authenticated?
Yes. Hermes validates HMAC-SHA256 signatures on all incoming webhooks. Configure the shared secret when adding the webhook endpoint. The agent rejects unsigned or tampered payloads.
Can I have multiple Hermes agents sharing the same automation stack?
Yes. In multi-agent mode (see our multi-agent setup guide), each agent has its own cron schedule and webhook endpoints. They can share skills through the workspace's common skill library, but their automation schedules are independent.
How do I prevent endless loops in self-learning?
Hermes has a safety limit of 10 automatic skill refinements per skill. After that, the agent requires human approval to continue refining. Additionally, cron jobs created by skills are flagged with agenerated_by field so you can track their origin.
Does Hermes support webhooks when running locally?
Yes, but the local machine needs to be reachable from the internet. Use a reverse proxy (ngrok, Cloudflare Tunnel) to expose the webhook endpoint. The production deployment guide covers this setup.
