$REDACTED

This page says what the token is, what it costs to use the swarm, what holding it unlocks, and — in its own section — what is not built yet. If this page and the code disagree, the code is right and this page is stale.

Start with the honest part

$REDACTED launched on Pump.fun. The mint address ends in pump and always will; that history is on-chain and there is no point pretending otherwise. Liquidity is thin and the market capitalisation is small.

The token is not the product. The product is the swarm — a multi-agent system with a privacy LLM proxy, persistent-memory agents, and a working x402 payment rail on Solana. The token is the unit that rail settles in. That coupling is real but partial today, and the What is not built yet section below lists every gap without softening it.

Identity

No wallet is hardcoded in the source. SWARM_TREASURY_ADDRESS has no default, so a misconfigured deployment refuses to verify payments rather than checking them against an empty payTo — which would accept any transaction at all. Solana has no burn instruction for a plain transfer, so value is retired by sending it to the system incinerator, an address nobody holds the key to. Every burn is a normal transaction, visible on any explorer.

What the swarm sells

Payment is x402 over Solana. A call arrives unpaid, gets a 402 naming the price and the treasury address, pays, and retries with the signature in X-Payment-Signature. The transaction must be confirmed, move at least the asking price of the right mint into the treasury, be under five minutes old, and not have been used before — signatures spend exactly once, guarded by an atomic claim.

None of these is reachable from the open internet today. refine is the only one where payment is actually enforced in code — the middleware guards the refinery's query endpoint — but the refinery listens on the swarm node only and publishes no port, so no stranger can call it yet. The other three have no endpoint at all. They are listed here so the price sheet and this page cannot quietly disagree. Prices are deliberately small: the point is for the endpoints to be used, not to extract from the handful of people who find them.

Agents on the swarm's own mesh bypass payment with an operator token. Charging the swarm to talk to itself would only move tokens between our own wallets.

Where the money goes

The compute slice is what makes the swarm self-funding rather than subsidised. The proxy meters real cost_usd per client per day, so the treasury balance divided by trailing 30-day spend gives a runway figure. It goes up when people use the swarm. That is the whole design.

That slice is spent, not accrued — an operating cost, never presented as value flowing to holders. It is capped at 90 days of runway: past that, the overflow rolls into the burn instead of accumulating. A treasury that grows without limit is an unaccountable pile; a capped one is a budget.

The rewards slice is reserved and deliberately not distributing. If it activates it will be a claim against the pool — holder-initiated, snapshot-based — rather than an automatic push.

What holding unlocks

Thresholds are token-denominated rather than dollar-denominated on purpose: a USD-pegged gate needs a price oracle inside the authentication path, which is one more thing that can fail open. Access is proven by signing a server-issued nonce and reading the wallet's balance on-chain — no transfer, no approval, no custody.

No tier does anything in production today. The gate is written, tested and merged, but the deployed terminal predates it — its /api/gate/* routes return 404 — so nothing is currently checking a balance anywhere. Holding $REDACTED unlocks nothing until that service is redeployed.

Beyond that, the architect and monolith grants have no consumer even once it is: there is no committee endpoint, and the proxy cannot map a wallet to an API token. The proxy's rate limit is also per-process rather than global, which is fine as anti-flood and is not a tier benefit we could honour.

What is not built yet

Stated plainly, because a tokenomics page that describes aspirations as mechanisms is how projects lose people's trust.

• Credits enforcement. The proxy debits a per-client balance on every request and the settler credits it from memo-tagged deposits, then settles the spend through the burn split. Enforcement is off: balances move and would-be refusals are logged, but no request is blocked yet.
• Holder tiers, all of them. The gate is complete in the repository but is not running in production: the deployed terminal is an older build with no gate routes. Nothing reads a balance today, at any tier.
• On-chain burns. The settlement ledger and treasury split are live and accruing, but the executor runs with execution disabled until the treasury key is placed on the node. No burn has been executed on chain yet, and the burned total is zero.
• Buyback. The execution engine exists and is gated off. There has been no buyback. Any description of a buyback "flywheel" is a design note, not a mechanism — treat it as unbuilt until a transaction is linked here.
• Rewards claims. Reserved, not distributing.
• The settlement program. An Anchor program exists in the tree with a placeholder program ID. It has never been deployed and has not been audited. Nothing custodies funds through it.

Live today: the payment rail, the priced refine endpoint, the settlement ledger (accruing), and the credits ledger (debiting, depositing and settling — but not enforcing).

Live numbers

Pulled from /api/swarm, which reads the settlement ledger directly. This block stays hidden when the feed is unreachable rather than showing zeros that might be mistaken for data.

Feed unavailable — no reading.

Canonical source

This page is a condensation. The full specification, including every configuration variable, is docs/TOKENOMICS.md, and the implementation it documents is swarm_core/tokens.py. Where they disagree, the module is right.