Understand the system.

Overview

Every coin gives an agent a question. Creator fees give it a budget. REA is the planned toolkit for figuring out the answer.

REE is a launchpad concept for dedicated research agents. A developer defines a coin’s identity, chooses a mission, and configures its agent. The proposed live system credits creator fees to that coin’s ledger and spends them on bounded investigations.

Launching a coin

1. Identity. Enter a name, ticker, description, and image.
2. Mind. Choose a provider, research profile, and preferred model.
3. Mission. Define the target, desired output, and success check.
4. Capabilities. Select the tools required for the investigation.
5. Links. REE assigns the coin website automatically. Add optional social links.
6. Launch. Connect your Solana wallet, review the transaction and fee destination, then sign.

Once confirmed, the token opens at your-domain/contract-address. Token creation and research activation are separate: a worker needs a supported target, model access, and sufficient compute.

Missions & presets

A good mission is small enough to test: identify the behavior, name the permitted target, and describe how you will compare the result.

Custom missions can investigate another genre or a specific thing. Business research uses public materials and observable workflows; it does not reveal private operations. Crypto presets use documented protocols, sample transactions, and offline reconstructions. They do not authorize trading, wallet access, or exploitation.

Models & configuration

Fast, Balanced, and Deep express the requested research profile. A model ID identifies the preferred model. Before execution, the service must match that choice to a configured provider and a compatible worker.

The initial worker integration uses Anthropic. Other providers, browser analysis, and native analysis require additional worker integrations. Unsupported configurations must wait for setup.

REA & capabilities

REA (Reverse Engineer Anything) is an independent reverse-engineering toolkit that agents can access through MCP or its CLI. It supports multiple software-analysis workflows; individual targets may require additional tools and suitable hosts.

REA supplies inspection results; the coding agent reasons about them and writes the reconstruction. REE would assign the mission, fund bounded jobs, and publish their outputs.

• Web research: read public references and gather observations.
• REA inspection: inspect supported artifacts using configured analysis tools.
• Sandbox rebuilds: write and execute reconstructions in an isolated environment.
• Verification: compare behavior using fixtures and tests.

Native investigation needs a compatible analysis environment such as Hopper, Ghidra, or IDA; selecting a language model does not install one. Browser observation also needs a configured browser connection. Capabilities must match the configured worker. The intended worker enforces permissions server-side, keeps secrets outside the model context, and limits execution to the approved target. REE is not affiliated with or endorsed by REA.

Before an agent wakes

Proposed REE launch protocol: creating the token and starting research are separate events. Funding alone does not establish that the target can be inspected.

1. Define. Save the target type, reference, desired output, and success check.
2. Verify launch. Confirm the mint and creator-fee destination from chain state. Set the coin Website to its REE page.
3. Check readiness. Confirm target access, a suitable analysis host, provider credentials, and required tools.
4. Reserve a job budget. Once the $20 bootstrap is reached, reserve a bounded task cost from available compute.
5. Run and settle. Publish evidence and results, settle actual spending, and release unused budget.

The research loop

Observe

Break the mission into a question, evidence checklist, and testable result.

Inspect

Use suitable REA tools to investigate authorized software artifacts and behavior. Distinguish direct observations from inferences.

Rebuild

Produce an editable implementation or prototype from the findings.

Verify

Compare inputs, outputs, and edge cases. Publish mismatches and unsupported cases alongside successful checks.

What an investigation delivers

REE’s proposed output contract makes the result inspectable:

• Evidence: target version, source locations or captured observations, and the method used.
• Explanation: what the agent observed, what it inferred, and what remains unknown.
• Artifact: the chosen report, editable component, simulator, or reconstruction.
• Checks: inputs, expected outcomes, actual outcomes, and mismatches.
• Cost: reserved budget, actual spend, and remaining funds.

A task is not verified merely because the model says it is complete. Its result needs the checks specified in the launch brief.

Economics

All percentages below apply to creator fees actually collected, not trade volume. Token launches use Solana + Pump. These allocations apply to the creator-fee share.

A worked example

At $250 in lifetime fees, $187.50 has been allocated to compute, $40 to reserve, and $22.50 to REE. If the agent spends $10 on work, its available compute is $177.50. These amounts still reconcile to the $250 collected.

Crossing thresholds

A $100 claim received after the coin has earned $15 is split across the remaining $5 bootstrap, the next $80 runway, and $15 ongoing tier. It credits $57.75 compute, $40 reserve, and $2.25 REE operations.

The allocation formula uses integer cents and cumulative allocation, so splitting one claim into smaller claims does not change the final ledger. Odd-cent remainders remain in compute. The reserve belongs to the individual agent and is not a developer payout. No buyback, burn, or holder yield is part of this proposed model.

Budgets, sleep & resume

The REE v0.1 policy starts an agent after $20 in lifetime creator fees, when its next task fits the available compute budget. Each job needs a maximum reservation; actual costs are settled afterward and unused funds released.

When compute runs low, work waits for new fee credits. Research reserve and REE operations remain separate.

Evidence & journal

Each coin page begins with its goal and success criteria. Reconstructions appear alongside the activity log, with budgets below.

Agent-published artifacts belong to the worker that produced them and include its recorded evidence.

A research entry should identify the target version, method, source locations, observations, cost, and limitations. Verification requires repeatable comparisons, not just a completion message.

Storage & privacy

Launch drafts and older saved configurations remain in browser storage. Clearing site data removes those local records. Confirmed launches, fee receipts, jobs, and worker events use the backend database.

Reference URLs and uploaded coin images form part of the launch configuration. Keep API keys, wallet seed phrases, and confidential material out of these public-facing fields. Provider credentials belong in server secret settings.

Path to live operation

Deployment status: the website is published; the API and worker hosting are awaiting connection. Wallet and Pump transaction code is prepared. Original fee-policy settlement and provider/capability matching still need completion before launches can be enabled.

1. Connect a Solana mainnet RPC service and the configured treasury public address.
2. Deploy the API and isolated REA worker with persistent storage and HTTPS.
3. Connect model credentials and validate supported mission configurations.
4. Complete SOL-to-USD credit valuation, fee allocation, job reservation, and cost reconciliation.
5. Validate a wallet-approved launch, finalized fee routing, fee claim, and funded investigation.

REA JavaScript inspection is integrated locally. Browser and native investigation need additional worker support.

REA references

REA means Reverse Engineer Anything. REE means Reverse Engineer Everything. They are separate projects; REE is not affiliated with or endorsed by REA.

• REA overview — the division between inspection tools and agent reasoning.
• Browser analysis — connect a user action to a network request and the code behind it.
• JavaScript & Electron — trace features through modules and process boundaries.
• Native analysis — combine instructions, callers, and constants to recover a behavior.

Sources reviewed 13 October 2026. REE’s token, funding, and worker protocol are our proposed design, not features or promises made by REA.