Let Claude Code invoke itself.
Run agents on schedules, chat with them on Discord, and resume any session in your terminal.
Documentation • Getting Started • Discord • Issues
Claude Code has changed the world, but most of the work it does is supervised by a human. What if Claude Code could invoke itself?
herdctl is an orchestration layer for Claude Code. Under the hood, it uses the Claude Agents SDK to trigger Claude Code sessions on schedules, via webhooks, or via chat messages.
It uses simple .yml files that are a thin wrapper around Claude Agents SDK configs, plus a couple of extra features provided by herdctl - schedules, chat and hooks. Each Agent defined this way can have any number of schedules that trigger the agent to do something automatically.
herdctl can run fleets of agents, each with their own source-controllable configurations. Agent fleets can be run anywhere via a simple herdctl start command.
herdctl allows you to interact with all of your existing Claude Code projects via chat apps like Discord and Slack (Telegram support coming soon).
Self-Invoking Agents — Define agents that wake themselves up on schedules or triggers. Coordinate an entire fleet from a single herdctl start command.
Fleet Composition — Build "super-fleets" from multiple project fleets. Each project keeps its own agent configurations, but they all run together with a unified web dashboard and CLI.
Full Claude Code Power — If Claude Code can do it, your herdctl agent can do it. Same tools, same MCP servers, same capabilities. herdctl is a thin orchestration layer, not a sandbox.
Two Runtimes — CLI runtime uses your Claude Max subscription (much cheaper per token). SDK runtime uses API pricing. Both support Docker isolation with resource limits and network controls.
Chat From Anywhere — Connect agents to Discord or Slack. Message your agents from your phone, get responses, and they continue working based on your conversation. Your PR reviewer bot becomes a team member you can @ mention.
Session Continuity — Every job creates a real Claude SDK session. When an agent finishes, you can claude --resume that exact session in your terminal. Pick up where the agent left off with full context intact.
Bidirectional Communication — Agents write structured data back to herdctl via metadata files. Hooks act on that data. Coming soon: agents that request schedule changes, store persistent context, and evolve their own behavior over time.
In which we try to cover all of the main parts of herdctl in 20 minutes:
Watch the demo
# Install herdctl globally
npm install -g herdctl
# Initialize a new project with example agents
herdctl init
# Start your agent fleet
herdctl start
Your agents are now running. Check their status:
herdctl status
Fleet overview showing agents, status, and recent jobs
The web dashboard gives you a browser-based control panel for your fleet. Enable it with herdctl start --web or web.enabled: true in your config.
See the web dashboard documentation for full details.
| Package | Description |
|---|---|
herdctl | CLI for fleet management. Install globally, run herdctl start. |
@herdctl/core | Core library. Embed fleet management in your own applications. |
@herdctl/web | Web dashboard. Real-time fleet monitoring, agent chat, and job management in your browser. |
@herdctl/discord | Discord connector. Chat with your agents via Discord DMs and channels. |
@herdctl/slack | Slack connector. Chat with your agents via Slack channels and DMs. |
@herdctl/chat | Shared chat infrastructure. Session management, streaming, and message handling used by all connectors. |
Full documentation at herdctl.dev:
Contributions welcome! See CONTRIBUTING.md for development setup and guidelines.
MIT License - see LICENSE for details.
Built by Ed Spencer
FAQs
Autonomous Agent Fleet Management for Claude Code
We found that herdctl demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Multiple high-impact npm maintainers confirm they have been targeted in the same social engineering campaign that compromised Axios.
Security News
Axios compromise traced to social engineering, showing how attacks on maintainers can bypass controls and expose the broader software supply chain.
Security News
Node.js has paused its bug bounty program after funding ended, removing payouts for vulnerability reports but keeping its security process unchanged.