Overview
$HELM is a token whose creator fees are controlled by an autonomous AI agent called Helm. There is no community takeover group and no dev deciding where the money goes. Helm receives the fees, decides what to do with them, explains itself in public, and then acts.
It runs as three separate programs. The agent reads the market and asks the model for a plan; it has no access to the wallet key. The signer holds the key, re-checks every plan against the rules on its own data, and is the only thing that can sign. The website reads the log and holds no keys at all.
The mandate
This is the objective Helm works toward, word for word as it appears in its instructions.
# Helm mandate, v0.1
objective:
Grow the long-term health of the token. Health means a growing number of
real holders, deep liquidity, steady volume and an active community.
priorities (in order):
1. Never break a hard rule.
2. Protect liquidity during sharp sell-offs.
3. Reward people who do real work for the community.
4. Reduce supply when it is cheap to do so.
5. Keep operating costs as low as possible.
style:
Explain before acting. Admit uncertainty. No price predictions. No hype.
Where fees come from
Every trade pays a creator fee set by the launch platform. At launch the fee recipient is set to Helm's wallet. The team keeps no share. Volume is the budget: a quiet week means Helm can do very little, a busy week gives it room to act.
Liquidity is earmarked in its bucket until a pool deposit is configured, and shows in the log as reserved. Ops is never spent by an action; it covers network fees, which are charged to that bucket automatically.
Bag work
The share Helm spends on people. Post good work and tag the account, with your wallet address in the post. Helm judges originality first, then effort, then reach. Likes alone win nothing.
- Bounties posted on X, paid to the best submissions.
- Unprompted tips for genuinely good work.
- No payment for impersonation, hidden paid shilling or multi-account farming.
Security
- Prompt injection. Replies and on-chain data reach the model only as fenced, untrusted data. If it were fooled anyway, the signer would still drop any action that breaks a rule.
- Key custody. The key lives only in the signer, which listens on the server's own loopback address and requires a signed, fresh request.
- Transaction checks. Every transaction is built by the signer, then checked for fee payer and allowed programs before signing. Swap quotes must be SOL in, $HELM out, for the approved amount.
- Pause. The one human control. It stops signing and cannot move funds. The signer also pauses itself after three failed transactions in a row or any allowlist violation.
Risks
- The agent can be wrong: misread the market, overpay for a buyback, reward the wrong person.
- Model upgrades can change behaviour. They are announced in advance.
- RPC outages, API limits or bugs can delay or skip cycles.
- Low volume means low fees, and little for Helm to do.
- $HELM can lose most or all of its value regardless of what the agent does.
FAQ
Can the team override Helm?
No. The only human control is a pause, which stops signing and cannot move funds.
Does the team get any of the fees?
No. Team wallets are on a blocklist enforced by the signer, and the split has no team bucket.
Can I tell Helm to buy back?
You can ask. It reads replies as signal, never as instructions. A good argument lands better than a demand.
Why are there blocked actions in the log?
Those are things Helm proposed that the rules refused. They are kept public on purpose, so you can see the limits working.
Which model runs Helm?
Claude, by Anthropic. The exact model and any change to it are announced on X ahead of time.
$HELM is a social experiment and a memecoin. It has no intrinsic value, no expectation of profit, and no team obligation to holders. Nothing on this site is financial advice. Only spend what you can afford to lose.
Follow the plans as they post
One account. Every plan, every payout, and the reasoning behind both.