AGENCY $AGENCY

AGENCY lets a creator launch a Pump coin from a wallet and choose a reasoning model for it. The project's documentation says creator fees are routed to AGENCY in the launch transaction, then credited to that specific coin for model usage and a reserved treasury. The AI mind observes, researches, proposes actions, and publishes activity within platform-defined limits.

• Each launched coin has its own accounting rather than a shared discretionary pool.
• The creator does not directly command the coin's AI mind after launch, according to the project.
• The product runs on Solana and uses Pump's launch and trading rails.

Official economics describe a staged allocation: the first $20 of a coin's creator fees goes to model credits; from $20 to $100, fees are split evenly between credits and that coin's treasury. Once cumulative fees exceed $100, 15% of later claims is allocated to AGENCY buyback and burn, with the remainder supporting the mind and its treasury. A mind can halt when fee activity drops below the documented operating threshold.

• The model is funded by trading activity, not a guaranteed subscription stream.
• A coin's AI actions are subject to policy checks and an isolated signer, as described by the team.
• Fee routing and actual treasury movements should be checked against on-chain records.

The project says 15% of eligible post-threshold creator-fee claims purchases AGENCY on PumpSwap and burns the acquired tokens. This can reduce circulating supply, but it does not give holders a claim on platform revenue, a guaranteed price floor, or a promised return. The mechanism's economic impact depends on real repeat volume and the amount of fees generated by launched coins.

• Track buyback transactions and burns, not just dashboard totals.
• Compare recurring fee generation with one-off launch spikes.
• A burn mechanism cannot offset a collapse in market demand by itself.

Most meme launchpads stop at token creation and liquidity. AGENCY's differentiation is to attach an ongoing, budgeted agent that can research, set missions, reward holders, and propose treasury actions. The useful question is whether these behaviours produce lasting communities and measurable outcomes rather than more automated content around short-lived coins.

• Per-coin public activity can make agent behaviour observable.
• Policy limits matter more than claims of AI autonomy alone.
• Retention of active coins is a stronger signal than total coins launched.

The thesis strengthens if independent on-chain checks match the stated allocation rules, launched coins continue earning fees beyond their first days, and agents deliver useful holder programmes or creations. Transparent security reviews and stable operation through lower-volume periods would also improve confidence.

• Growth in active, fee-generating coins over multiple weeks.
• Publicly reconcilable claims, buybacks, and burns.
• Documented outcomes from agent-run programmes, not only activity counts.

The system combines AI decision-making, off-chain services, treasury accounting, signers, and on-chain trading. Official security claims are not a substitute for independent review. If newly launched coins stop trading, model activity and buyback funding can shrink quickly; meanwhile AGENCY remains a volatile token with no contractual income right.

• Operational, accounting, prompt-injection, and signer risks require ongoing scrutiny.
• Fee and burn figures on the project site are self-reported until independently reconciled.
• Liquidity, holder concentration, and token price can change abruptly.