Claude Code · Codex · invite-only beta

A map of your repo your agent and team share

getmyorg maps your code into features, keeps the map current as you commit, and guards risky commands in Claude Code and Codex.

Set up in a few minutes
Session guardrails
70+ specialist roles

Coding agents start every session cold.

They re-learn your repo each time, drift from your conventions, and will ship band-aids or force-push to main if you let them.

One assistant can't hold the structure of a real organization

Context, standards, and judgment of an entire org don't fit in one prompt window. Agents drift, duplicate work, and miss craft conventions.

Agents will ship band-aids if you let them

Agents suppress errors to make tests pass, disable failing tests, add fallbacks that hide real bugs, and bypass pre-commit hooks. Review density is too low to catch all of it.

Human-org defaults break in agent orgs

The cost ratios that justify shortcuts in human orgs invert when agents iterate at speed. Band-aids compound silently. Memory of 'we patched this' evaporates between sessions.

Nothing carries over between sessions

Each new session re-reads the repo and guesses at the architecture: which features matter, what's in flight, what was decided last week.

Up and running in a few minutes

Install and sign in once per machine, then map each repo. No configuration.

Once per machine
01Invite-only beta

Install

Anyone can install getmyorg; signing in needs an invite.

> /plugin marketplace add https://install.getmyorg.com/marketplace.json
then > /plugin install myorg@getmyorg

Then turn on auto-update for the getmyorg marketplace (/plugin → Marketplaces); it is off by default.

02

Sign in

Sign in with your invite. Your sign-in renews itself.

> /myorg:login
Then, in each repo
03

Map your repo

Maps the repo, installs the git hooks that keep the map current, and builds the semantic map once.

> /myorg:init
04

Look around

The map in your browser: the features, the code by system, and a Guide to what's next.

> /myorg:hq
05

Chart the rest

When you want the whole map, charts the code the first pass didn't explain. Large repos take longer.

> /myorg:semantic refresh --chart
claude-code — louislam/uptime-kuma
>/myorg:triage
Triage — 736 open issues (200 most recently updated shown)
Duplicates (15 groups)
[high] canonical #4771 "`Check Now` or `Test` Button to initiate a check before the regular time"
duplicate #7557 "Add check availability button on each monitor"
Why: both describe the same workaround: "saves people from pausing and resuming" vs "I have to pause and then resume each".
…
Ranked (200 issues)
Now (6)
…
#8181 keepDataPeriodDays default 180 vs 365 -- Silent data loss: saving any settings page on a fresh install shortens retention. The cause is pinned to Settings.vue:176 vs clear-old-data.js. Small, scoped fix.
#8193 Interval maintenance (>=2 days) fires once -- The maintenance window never recurs, so users get alerts they expected to be suppressed. Root cause pinned in server/model/maintenance.js.
#7956 Docker restart loop reported as UP -- Missed outage, which is the worst failure for a monitor. 6 comments, the reporter confirmed a fix, and a smaller fix (State.Status "restarting") is proposed.
#6025 No UP notification after DOWN->PENDING->UP -- Several reports (#5742, #4262, closed dup #7968). Still happens on 2.5.6. Recovery alerts are lost for docker, webhook and push monitors.
#8180 AlertNow: "[undefined]" / UP never sent -- Exact lines cited (alertnow.js reads heartbeatJSON.name). Quick win, but only affects AlertNow users.
…

Read-only: uptime-kuma's open issues, deduped, grouped and ranked. You apply the changes.

Real output on the public louislam/uptime-kuma repo at e6ea9509, with getmyorg 0.5.18; … marks a cut. Not affiliated with the Uptime Kuma maintainers.

See getmyorg's map of real repos →

See the map getmyorg drew of a real repo.

uptime-kuma, a popular self-hosted monitor, mapped by getmyorg: what the code does, grouped into features, with open issues placed on them.

How much of the code the map has read, and what it found.
HQ at phone width

getmyorg's map of louislam/uptime-kuma at commit e6ea9509, captured with getmyorg 0.5.18 on 2026-11-17. Not affiliated with the Uptime Kuma maintainers. Walk through this map →

Always in view, in Claude Code's status bar

How much of the code is charted, your folder, model and context, and the one thing the Guide says needs you next.

▦ getmyorg │ Opus 5.5 (high) · ○✓
Calm. Nothing needs you, so the right side is a check.
Opus 5.5 (high) · ○› /myorg:anchor
Needs you. Something does: the Guide's next step, in its severity colour.

Real output of getmyorg 0.5.21's status bar in a local copy of louislam/uptime-kuma at e6ea9509. The calm bar needs a goal merged to the default branch, so one taken from its README was merged in the copy first. Not affiliated with the Uptime Kuma maintainers.

A shared map, and guardrails that run today

What getmyorg adds to every session in Claude Code and Codex.

Always current

Project map

Your repo's features, code map and Guide, in your browser with /myorg:hq. Git hooks keep it current as you commit, switch branches and pull.

In git

One map for the team

The semantic map is committed with your code, so a clone or a pull loads the same map for everyone.

Before every edit

Edit checks

Before an edit lands, the band-aid detector scans what it adds: swallowed errors, lint disables, a bare @ts-ignore, skipped tests. The agent is asked to fix the cause or say why.

4 levels

Command guards

Risky shell commands are stopped or sent back for a second look, and a force-push to main is stopped at every level except off. You choose the level with /myorg:config guards: strict, guarded, relaxed (the default) or off.

Two guardrails running in your repo today.

Both install with the plugin, in Claude Code and Codex (in Codex, approve them in /hooks first). Here's what each one catches, what it says, and what it doesn't claim. Guardrails for agents, not a security boundary: you set the level, including off. What works today →

Band-aid detector

Checks what an edit adds before it lands.

before Edit(auth.ts) — adds // @ts-ignore
getmyorg guard note: this edit adds a pattern that often hides a problem instead of fixing it. If it's deliberate, add a short comment saying why (and an issue link for a TODO). Otherwise, fix the cause.
 
[getmyorg] EDIT NOTE (rule: deny.ts_ignore_naked)
matched: // @ts-ignore
sections: deny.ts_ignore_naked
level: /myorg:config guards ($myorg:config guards in Codex) shows or changes the guard level

Before an edit is applied, getmyorg scans the lines it adds for patterns that hide a problem instead of fixing it: swallowed errors, file-wide lint disables, a bare @ts-ignore, skipped tests, unexplained TODOs. The agent gets a note asking it to fix the cause or say why.

A note at the default level; the edit goes ahead. Only strict stops the worst matches. It doesn't re-check edits after they're applied.

Force-push gate

History other people build on stays put.

git push --force origin main
getmyorg's guard stopped this command.
 
[getmyorg] BASH STOPPED (rule: force_push.protected_branch)
command: git push --force origin main
why: force-pushes or deletes main, which other people's work builds on
level: /myorg:config guards ($myorg:config guards in Codex) shows or changes the guard level
 
You can't clear this stop yourself, and retrying won't help. If the command is really needed, explain why to the user. They can run it themselves with `! <command>`, or change the guard level with /myorg:config guards ($myorg:config guards in Codex) (guarded now). Don't change the guard level yourself.

A force-push or delete aimed at main, master or the default branch is stopped at every level except off, in Claude Code and Codex. The agent can't clear it. You can run the command yourself with ! <command>.

No evidence check or audit trail yet. Deploy and migration gating isn't built.
not a marketplacenot LangChain — agents already existnot observability — it tells you what agents did; we gate what they can dohonest about what's not yet built
Ask another agent

Get a second opinion from another agent

Ask another coding agent you have installed for a bounded answer or a critical review, then fold its answer into yours.

How it works
  1. 01Your agent writes a self-contained brief. Peers don't see your session's history.
  2. 02Other agents run on your own accounts, so it only calls them when you ask.
  3. 03It reads each answer and decides what to keep. A peer that hangs times out.
> /myorg:second-opinion "Review this migration plan"

In Codex: $myorg:second-opinion

Or ask one agent by name
  • /myorg:codex
  • /myorg:gemini
  • /myorg:opencode

Each one needs that agent's own CLI installed and signed in.

An org of specialists, one request away

70+ specialists across engineering, product, design, data and more. Ask with /myorg:dispatch and it routes the work.

Engineering

Building, reviewing, testing and debugging.

Product

Specs, priorities and what to build next.

Design

Flows, interfaces and UX reviews.

Data & Analytics

Pipelines, models and metrics.

What's encrypted, what's shared, what leaves your machine

getmyorg never uploads your code. Here is what it encrypts, commits to your repo and sends.

Encrypted

Your org's agents, skills and standards arrive as an AES-256-GCM encrypted bundle and stay encrypted in the local cache. The key is never written to disk.

In your repo, in plain text

The semantic map (.myorg/semantic-map.jsonl) and .myorg/LLM.md are committed so your team shares one map. The map database and session logs stay local and gitignored.

Sent to getmyorg

Sign-in (OAuth with PKCE), token refresh, an entitlement check and the bundle download. Your session token is kept in a file only your user can read.

Never sent

Your source code, your issues and your project map. Map analysis runs in your own Claude Code or Codex session.

Map your repo in a few minutes.

Install, run /myorg:init or $myorg:init, and open /myorg:hq or $myorg:hqto see your repo's features and map.

Honest about what's not built yet: what works today.

FAQ

Questions, answered plainly

No hand-waving. If something isn't built yet, the answer says so.

Read the docs