ConnectOnion is an open-source, template-first toolkit for building, debugging, deploying, and operating AI agents. Developers use its command-line tools and Python runtime to create agents, add tools, connect services, deploy them, and make them callable by other agents, while the catalogue entries are related agents, skills, and instructions.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/openonion/connectonionWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/openonion/connectonion/rem_help)<a href="https://agentmods.dev/commands/openonion/connectonion/rem_help"><img src="https://agentmods.dev/badge/commands/openonion/connectonion/rem_help/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/openonion/connectonion/rem_help"><img src="https://agentmods.dev/badge/commands/openonion/connectonion/rem_help.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00000 | $0.07571 |
| Opus 5.5 | $0.00000 | $0.03028 |
| Sonnet 5.5 | $0.00000 | $0.01514 |
| Haiku 4.5 | $0.00000 | $0.00757 |
Grade B, and why
rem_help scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Enumerates other installed skillsmediumAgent snooping
Other skills' SKILL.md files reveal prompts, capabilities and secrets that should be invisible to peers.
Example: co rem map-skills --skills-dir ~/.claude/skills How it starts
The opening of the file, as written. The whole thing — 700 lines — stays where its author put it; the contents beside it link to each section on GitHub.
co rem help pages
Each section is printed verbatim by --help for the command it names.
The agreed design is issue #1656; change a page there and here together.
co rem
co rem — a notebook about the people, projects and tools in your work, kept up to date from your mail and coding sessions.
Experimental: a preview; its commands may change before 1.9.0.
Build (once)
init Build the notebook from 90 days of mail and sessions, then write your page.
investigate Fill a page, a whole category, or your own page, using a model.
Read
open Browse the notebook in your browser.
list List pages in a category (people, projects, orgs, skills).
show Print one page.
search Find text across pages. No model.
Keep it current
start Approve sources and turn on the daily schedule.
stop Turn the schedule off. Pages and approvals stay.
status Is the schedule on, when it runs next, what ran today.
sync Run one update now.
Settings
sources Which mailboxes, coding tools and chats the notebook reads.
config Model, schedule times and limits.
logs What each run did, and what it cost.
doctor Check that everything the notebook needs is installed.
Options (before the command):
--root DIR The notebook folder (default ~/.co/rem).
--json Machine-readable output: {"ok", "data", "next"}.
First time: co rem init
Example: co rem search "term sheet" --in people
Every page: co rem <command> --help
Advanced: co rem advanced --help (scan, map-skills, stub, reflect, reflections,
propose, review, abstract, capture, projects)
Old names: unfinished, people, daily, subscriptions, subscribe, unsubscribe, route
and usage still work until 1.9.0 and print their new name.
co rem init
Build the notebook and investigate what matters now. One command.
First the map is made without a model: a page for each person you write to, each organization,
each coding project and each installed Skill, plus your own page, titled with your
name and filled with who you write to most and where you work. init prints that
page when the map is done. It lists 90 days of mail headers, the preview line your
provider lists with each message (to name people by your greeting), saved
contacts, and session metadata, then saves a private copy of each listed message
body, once, so investigating a person later reads it from disk.
Then it writes your own page by itself from everything you sent
and your coding sessions of the last 30 days (co rem investigate me): a quick
first pass in about 4 minutes, then the whole page alongside the rest. It also
investigates eligible people, those active in the last 14 days first (up to two years of
their mail), projects (recent first), organizations linked to those people,
and installed skills from their source instructions and retained run evidence,
12 pages at a time. The result should let you recognize useful relationships
and work immediately, with evidence cited on each page.
--first-people, --first-projects, --first-orgs and --first-skills cap a kind (0 for none).
Before the first page it says one total: about how many pages, ~how many billed
input tokens on your plan and ~how many minutes, an estimate from the median of
this notebook's own runs (before there are any, measured defaults). It names the
runner and model. Around 20% of a weekly runner allowance is a target,
not a hard limit: the selected investigation finishes even if it uses more.
The configured weekly safety floor still stops new pages when measurable.
Ctrl-C stops it, says which pages were written, and
keeps the map and every page. It is skipped, with the reason, when the runner is missing
or signed out, when no mailbox gave an address of yours, or when your page was
already written. More people: co rem investigate people. More projects: co rem
projects write.
Usage: co rem init [--days N] [--mine ADDRESS[,ADDRESS...]] [--name NAME] [--mail gmail|outlook]...
[--no-mail-archive] [--investigate | --no-investigate]
[--first-people N] [--first-projects N] [--first-orgs N] [--first-skills N]
Example: co rem init --days 90 --name "Aaron Xie" --mine [email protected],[email protected]
Inputs: Connected mailboxes (co auth google, co auth microsoft) and local Codex /
Claude Code sessions. --mine adds addresses that are yours (commas, or
repeat it). Addresses that look like yours are listed on one line, with
one command that confirms the ones you keep.
Options: --investigate Explicitly request the default investigation.
--no-investigate Build the map only.
--first-people N Cap eligible people (default all selected; 0 for none).
--first-projects N Cap queued projects (default all selected; 0 for none).
--first-orgs N Cap related organizations (default all selected; 0 for none).
--first-skills N Cap installed skill investigations (default all mapped; 0 for none).
Output: Your page's facts and where it is; one progress line per stage on stderr
(every step in .state/init-progress.log); pages under ~/.co/rem (or
--root); private files under .state/: source-inventory.md and .jsonl
(what was listed, window by window), and mail/ (one body per message,
per-person and per-project indexes). A seven-day window at the
200-message listing cap is split until all of it is listed. Mail bodies
never go into a page and never leave this machine. Re-running keeps
anything written and reuses saved bodies.
Effects: Writes pages and private files (owner-only). Reads mail bodies unless
--no-mail-archive. Mailboxes it read are subscribed for the daily round;
nothing is read in the background until co rem start is approved. The
map costs nothing; the subsequent investigation uses the configured
model for the selected people, projects and organizations. No schedule.
Takes: About 10 minutes to map 90 days of two mailboxes; saving bodies takes
longer; investigation time depends on the selected pages and runner.
An interrupted run resumes where it stopped.
Next: co rem open (read your page), then co rem start (keep it current)
Back: co rem --help
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today Changed · +5 lines 3713f0cc57e8
- yesterday Changed · +5 lines 45b967d29b4a
- 2d ago Changed · +27 lines dca30b5c3ebb
- 3d ago First seen · 663 lines · 0 tokens per session scan B 174d700218c7
rem_help is a command published in the GitHub repository openonion/connectonion (1,491 stars, last pushed today), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 7,571 tokens. A static security scan graded it B with 1 finding (enumerates other installed skills). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-30.
Other commands, from other repositories
hep-graph
Build an Agentlas automation by describing it, list saved ones, or request a run.
hep-update
Update the installed Agentlas runtime and every host adapter on this machine.
session
Capture, list, rehydrate, and promote llm-wiki session context. Supports automated hook capture into HUB/.sessions plus explicit promotion into topic raw notes.
agentlas
Agentlas commands — one entry point for network, build, storm, call, and the rest.
agentlas-graph
Build an Agentlas automation by describing it, list saved ones, or request a run.
agentlas-orch
Set or show which model runs the orchestrator and which runs the workers.