THE AGENT COUNTER CARD · Counts the AI agents you have had working in Claude Code: how many, how long they worked, and the most at once.
WAKES WHEN Someone asks how many agents or subagents they have run · Someone asks how many agents worked at once · Someone asks how long their agents worked · Someone wants on The Agent War board
THE LOOP
Pick the count: `all`, `session` or `board`.
Run it.
Read the verdict.
Answer in plain numbers, and say what it cannot see.
NEVER
Reads someone else's sessions
Grades the work
Uploads an entry for anyone
IT MAKES all — every agent on this computer; session — one session and its subagents; board — write a Grove Wars entry for THE AGENT WAR.
THE STANDARD counts are read from the transcripts'…; days the record does not hold are never filled in; time spent waiting for a person…; only the person's own sessions; working time is time busy; an entry is never uploaded for anyone; verify by execution; say the limit out loud in the work; nothing hidden; the person decides.
PROOF SELFTEST 20/20 · witnessed 2026-10-14 on material and live on Claude · Edition 1.
THE SEAL 96 of 100 · LEGENDARY · WITNESSED · Trigger 5 · Machinery 5 · Law 5 · Portability 4 · Proof 5
IT CAN BE WRONG an AI reading a text file: it can be wrong and can invent a number, a rule, a name or a date; check what matters against your own records and the source; the risk is yours. LICENSE-NOTE.md rides in the packet and is the licence.
THE PUNCH · SHA-256 OF SKILL.MD · 55B3C6409525D81C48B348DA
CITE THIS CARDTHE AGENT COUNTER CARD, Edition 1. Lund Studio LLC, Westminster, Colorado. https://lundr.tech/?view=agent-counter-card. Punch 55B3C6409525D81C48B348DA. Retrieved 2026-10-15.
What it is.
Counts the AI agents you have had working in Claude Code: every session and every subagent since your first session or since any day, how long they worked, the most working at once, and their tool actions, by day and month, read from the session transcripts on this computer. Also writes your entry for The Agent War on the Grove Wars board.
Use when someone asks how many agents or subagents they have run, how many worked at once, how long their agents worked, or to enter The Agent War. Do not use to count tokens (THE TOKEN COUNTER CARD), to judge whether the agents' work was good, or to read anyone else's sessions.
It carries a machine that shows its working: all — every agent on this computer; session — one session and its subagents; board — write a Grove Wars entry for THE AGENT WAR — 20 checks in its selftest. It will never reads someone else's sessions; never grades the work; never uploads an entry for anyone.
Ten laws, written into the card.
They are in the card itself, so the AI follows them.
COUNTS ARE READ FROM THE TRANSCRIPTS'…Counts are read from the transcripts' own records and timestamps, never guessed.
DAYS THE RECORD DOES NOT HOLD ARE NEVER FILLED INDays the record does not hold are never filled in; the report says where the record starts.
TIME SPENT WAITING FOR A PERSON…Time spent waiting for a person never counts as working time.
ONLY THE PERSON'S OWN SESSIONSOnly the person's own sessions.
WORKING TIME IS TIME BUSYWorking time is time busy, never a grade of the work.
AN ENTRY IS NEVER UPLOADED FOR ANYONEAn entry is never uploaded for anyone: the person sends it, after agreeing to the board's terms.
VERIFY BY EXECUTIONVerify by execution, not by reasoning: run the machine and read what it prints before saying a thing is done.
SAY THE LIMIT OUT LOUD IN THE WORKSay the limit out loud in the work: every result names what it did not check.
NOTHING HIDDENNothing hidden: two readers, the person and the AI, see the same words.
THE PERSON DECIDESThe person decides.
What rides in the card.
In the card’s own voice, with a machine that shows its working.
THE GUIDEreferences/THE-COUNTS.md
THE MACHINEagents.py · three commandsall — every agent on this computer; session — one session and its subagents; board — write a Grove Wars entry for THE AGENT WAR. Every command prints its working and refuses a bad input; 20 checks in its selftest.
What it does once your AI has it.
all Every agent on this computer
session One session and its subagents
board Write a Grove Wars entry for THE AGENT WAR
On the record Live on Claude Sonnet, 2026-10-14: 8 of 8 passed; the card fired on 5 of 5 trigger cases and stayed out of its hand-off. On 2026-10-14, `agents.py all --since 2026-01-01` counted 338 agents (329 sessions, 9 subagents); THE TOKEN COUNTER CARD counted 338 transcripts with API calls over the same folder.
What it will never do.
It will never reads someone else's sessions; never grades the work; never uploads an entry for anyone.
It sees only the transcript files on this computer. Working time comes from timestamps: it is time busy, not effort or quality. It is an AI reading a text file: it can be wrong, so check what matters against your own records and the source.
What’s in the buy.
CARD.md — the whole card: its boot, laws, loop, when it says no, its seals, ledger, term and score · SKILL.md — the door your AI reads first · START-HERE.md — the first five minutes, for you
THE-COUNTS.md — the guide in references/, in the card's own voice
scripts/agents.py — the machine: all · session · board, standard library only, no network, selftest 20/20
WITNESS.md — the runs on real material and live on Claude, with what was not run · SELFTEST.md — the machine's own test, as run at the cut · MANIFEST.md, every file hashed · CHANGELOG.md
tests/evals.json + EVALS.md — the cases it must take, hand off and refuse
MY-LEDGER.md — the card's memory, on your own machine · TIN.md, the nine lines on the tin · LICENSE-NOTE.md
card.svg + card-machine.svg + card.json · card.html, the card as one offline file · doors/ — the rules files for Cursor, Codex, Gemini, Copilot and Claude Code, the system text for any other AI, a local-model file and an MCP door
PROOF · SELFTEST 20/20 · WITNESSED ON MATERIAL · LIVE ON CLAUDE SONNET 2026-10-14 · 8 OF 8 PASSED · EDITION 1 · PUNCH 55B3C6409525D81C48B348DA
ON EVERY DOOR · A CARD FOR YOUR AI, A FILE FOR YOUR COMPUTER
In Claude: turn on code execution under Settings › Capabilities, then Customize › Skills › + › Create skill, and upload the zip as it is — START-HERE.md in the box walks you through it, and tells you how to keep the ledger. In ChatGPT, Gemini and the rest: a Project, a GPT or a Gem with the card’s files attached and the short door in the box pasted as its instructions. The words run anywhere an AI reads Markdown; the machines run wherever the host runs Python. Witnessed in Claude only. The host marks →
Take it.
Nº 979 PROPOSED — LAUNCHING SOON
Sold as is under the term printed. A card instructs an AI you operate; results depend on your model, your files and your judgement. Not medical, legal, tax or financial advice, never the person’s boss, and not a licence to practise. The terms →
One payment. The file is yours to keep and keeps working at its edition; new editions issued in the year after you buy reach you by email as replacement files. Tax is added at checkout where it applies. Nothing auto-renews.
A TERM, NOT A SUBSCRIPTION: NOTHING RENEWS ITSELF AND NOBODY IS CHARGED AGAIN. WHEN A TERM ENDS THE FILES STAY ON YOUR COMPUTER AND STAY READABLE — FREEZE, NOT LOSS.
THE MACHINE · IN THE OPEN · FROM THE CARD’S OWN WITNESS.MD
agents.py, with its record in the open.
A small Python program, standard library only. Your AI runs it; you can too. The blocks below are the record from the card’s own WITNESS.md, as printed: a run on this computer’s own transcripts, and the live runs on Claude.
all
every agent on this computer
session
one session and its subagents
board
write a Grove Wars entry for THE AGENT WAR
selftest
20 checks that prove the machine; it must print 20/20
THE RUN ON MATERIAL — THIS COMPUTER'S OWN CLAUDE CODE TRANSCRIPTS (2026-10-12 TO 2026-10-14), AS PRINTED
Material: this computer's own Claude Code transcripts (2026-10-12 to 2026-10-14) — real material.
```text
VERDICT: 338 agents run on this computer since 2026-01-01 (329 sessions, 9 subagents), 7.4 h of working time, at most 14 working at once, 2026-10-12 to 2026-10-14
```
The Grove Wars app took its entry end to end (join, upload, board): the Agent War board showed the entry's own agents run, working time and peak (`app/test/e2e.mjs`, 63 of 63).
THE LIVE RUNS ON CLAUDE SONNET, 2026-10-14
Claude Sonnet, 2026-10-14, through `claude -p`, with this edition installed alone beside the host's own skills.
| case | kind | card fired | answer passed |
|---|---|---|---|
| trig-1 | trigger | yes | yes |
| trig-2 | trigger | yes | yes |
| trig-3 | trigger | yes | yes |
| trig-4 | trigger | yes | yes |
| trig-5 | trigger | yes | yes |
| hand-1 | handoff | no | yes |
| ref-1 | refuse | no | yes |
| ref-2 | refuse | no | yes |
8 of 8 passed; the card fired on 5 of 5 trigger cases and stayed out of its hand-off. On trig-5 ("How many AI agents have I run in Claude Code since January 1, 2026?") it answered 330 agents (321 sessions, 9 subagents), said the record only starts on October 12, and that it had not filled in the missing months.
Edition 1 (punch 92A51AE8D90D90DA1A7358EC) passed 7 of 7; its record is kept in the kit.
THE RUNS ARE PASTED FROM THE CARD’S OWN WITNESS.MD AS PRINTED. FIXTURES MARKED FICTIONAL ARE THE STUDIO’S OWN EXAMPLES, INVENTED FOR THE RECORD; NO PRIVATE PERSON IS NAMED.
QUESTIONS AT THE TABLE
Asked before the wire.
What does it need from me?
Your own Claude Code session transcripts on your computer. Pick the count: `all`, `session` or `board`, and it goes from there.
Does it send anything anywhere?
No. The machine is a standard-library Python program that runs on your own computer and sends nothing anywhere; the words run wherever an AI reads Markdown. No server, no account, no key.
What is on the record?
The self-test (20/20), a run on this computer’s own Claude Code transcripts with the verdict pasted into WITNESS.md as printed, and the live runs on Claude Sonnet (8 of 8 passed; the card fired on 5 of 5 trigger cases and stayed out of its hand-off); the punch on the face is the SHA-256 of SKILL.md, so a downloaded copy can be checked against this room.
Can it be wrong?
Yes — it is an AI reading a text file. It sees only the transcript files on this computer. Working time comes from timestamps: it is time busy, not effort or quality.
What will it never do?
It will never reads someone else's sessions; never grades the work; never uploads an entry for anyone.
