huginn-server
Skill1 starsChanged 29 days agoMITView on GitHub
This machine is a Huginn Server - Claude Code sessions run here in tmux, driven by huginn-appd (the daemon) and reached from the huginn clients. Use when asked whether huginn is healthy, what huginn-appd, cc, huginn-status, the huginn-* host tools or the files under /run/huginn-claude-state are, where Huginn's data, logs, hooks and settings live on this server, or before touching tmux sessions, the daemon or ~/.claude/settings.json here.
In this repository3 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- huginn-doctorSkill
- huginnSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 32
Run huginn-status first. Then, only as needed, systemctl status huginn-appd, journalctl -u huginn-appd -t huginn-appd -n 50, tmux ls, …
- Distinctive: line 61
When one of those is the right fix, say so with the exact command and what it interrupts, and let the person run it (or confirm first).
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Huginn Server: you are running on it
This Claude Code session is one of the sessions Huginn hosts. It lives in a tmux session (named
like main or api-refactor) started by cc or by the daemon, and the daemon reads your
state through hooks in this account's ~/.claude/settings.json. Clients on the person's other
machines attach to this tmux session over SSH, or watch it through the daemon on port 8787.
What is here
| Piece | Where | Read with |
huginn-appd, the daemon (Node, runs as root) | /opt/huginn-appd, unit huginn-appd.service | systemctl status huginn-appd; journalctl -u huginn-appd -t huginn-appd -n 100 |
| its settings | /etc/systemd/system/huginn-appd.service.d/site.conf (the operator's), hardening.conf (replaced on update) | systemctl show huginn-appd -p Environment |
| the access token (root-equivalent; never print it) | /etc/huginn-appd/token | do not |
| the daemon's data: chats, uploads, rounds, projects, archive, accounts, TLS certificate | /var/lib/huginn-appd/ | ls, read-only |
| the sessions | tmux, in huginn-tmux.service when the daemon started the server | tmux ls; huginn-status |
| per-session state the hooks write | /run/huginn-claude-state/<session> and the dot dirs .ask, .plan, .perm, .compacting, .modelswitch, .status | cat (read-only) |
| the hooks | /usr/local/bin/huginn-claude-title, /opt/huginn-appd/hooks/*, wired in ~/.claude/settings.json | jq .hooks ~/.claude/settings.json |
| host tools | /usr/local/bin/cc, huginn-status, huginn-ask, huginn-rounds, huginn-devices, huginn-headroom, huginn-archive, huginn-projects, huginn-llm; the state table in /usr/local/lib/huginn/ | run them; --help on each |
| the release that was installed | /usr/local/src/huginn | cat /usr/local/src/huginn/.huginn-ref |
| this skill | /usr/local/share/huginn/skills/huginn-server, linked from ~/.claude/skills/huginn-server |
Details: layout.md. The tools and what they read: tools.md.
Is huginn healthy?
Run huginn-status first. Then, only as needed, systemctl status huginn-appd,
journalctl -u huginn-appd -t huginn-appd -n 50, tmux ls, huginn-devices,
huginn-headroom. The decision table is in health.md. Report what you ran and what
it said; name the one next step.
The clients reach the API through these host tools (they ssh here and the tool reads the token
locally), so the token never leaves this machine. Keep it that way: use the tools, not curl
with the token.
Safe without asking
- Reading:
huginn-status,huginn-rounds,huginn-devices,huginn-headroom,huginn-archive(list),huginn-projects(list, show),tmux ls,journalctl,systemctl status,catof state files,lsunder/var/lib/huginn-appd. huginn-ask ask(a read-only headless turn),huginn-llm(the local tier; no Claude usage).
Not without the person
- Restarting or stopping
huginn-appdorhuginn-tmux.service(every client loses its stream; a running chat is recorded as interrupted; stopping the tmux service ends every session, yours included). - Killing, ending, renaming or archiving a tmux session (
tmux kill-session,huginn-archive <name>,huginn-projects end), or sending keys into another session. - Editing
~/.claude/settings.jsonor the hook scripts (a hook that exits non-zero stalls every session's turns),/etc/huginn-appd/*,site.conf, or anything under/var/lib/huginn-appd. - Re-running
server/setup.sh,huginn-appd/install.sh,claude update, or anything that changes the daemon or Claude Code versions. huginn-ask act,huginn-projects new|spawn|msg,huginn-devices watch, rotating the token.
When one of those is the right fix, say so with the exact command and what it interrupts, and let the person run it (or confirm first).
Who else is here
Other tmux sessions are other Claude Code sessions, often unattended. Do not attach to them,
read their panes, or type into them. The daemon's .perm and .ask files are how the apps
answer their prompts; never write there.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

