Monitor your crypto portfolio with visual AI agents,survive the cycle.

No. Every workflow stops at a HITL gate before any signing path. The agents pull on-chain data, score concentration, draft the rebalance, and stage it for your approval. You can lift the gate per template once a routine has earned trust, but the default is review every move.

Three ways. Static context attaches your allocation policy, chain whitelist, and cost basis method to specific personas on every run. The rag_retrieve tool lets PortfolioManager and ChiefRiskOfficer pull from prior reports, protocol audits, and your DeFi position notes on demand. Cross-run memory means yesterday's drawdown reading is in scope for today's check. You choose which agent gets which layer.

It sits on top of them. Zerion and DeBank show you balances and recent moves across chains. Melaya is the reasoning layer that reads those positions through scoped RPC and indexer tools, scores concentration, runs the stress test, and writes the rebalance memo with citations.

Koinly and CoinTracking ingest transactions and produce tax forms. Melaya adds the reasoning layer on top. TaxStrategist work runs through PortfolioManager and ReportWriter, modeling FIFO, HIFO, and Specific ID across your wallets so harvest decisions land before year end, not after.

No. The Quant Firm crew is six agents and one canvas. A solo hodler runs the same three templates a fund runs: weekly risk check, monthly rebalance review, and year-end tax pass. You pay for what each agent thinks, not per seat.

Any. Claude on ChiefRiskOfficer where stress test reasoning earns the cost, GPT on ReportWriter for the one-page synthesis, a local Ollama on QuantAnalyst when wallet history cannot leave your machine. Each persona picks its own model.

With an Etherscan key, an RPC URL, and a CoinGecko quote feed authorized, the portfolio risk check workflow is a 4-node canvas: pull balances, score concentration, draft summary, stage for review. Most solo hodlers ship it the same evening.

They are not. Every blockchain tool in the bundle is read-only: eth_balance, eth_txlist, btc_address_info, sol_balance. No signing tool ships with the hodler templates, so the worst an agent can do is read your address and write a memo.

Every run logs every tool call, every RPC fetch, every model invocation, and every approval decision. Replay any run to see which address, which block height, and which price snapshot drove the conclusion. The audit log is the methodology log.

n8n, Zapier, and Make execute predefined trigger-action steps, and a hand-rolled script breaks the day an explorer API changes. Melaya's Quant Firm crew plans across steps on a canvas of agents, tools, triggers, and approval gates: the Quant Skeptic attacks the Portfolio Manager's thesis before the Report Writer ships an ACTION line. Zapier and Make still win on connector breadth for simple linear automations.

Yes. Save any run as a template and the crew, tool allowlist, and approval gates come with it, so the Chief Risk Officer's VaR and stress test pass runs the same way every week. Every run keeps full run traces with typed failure reasons and cost per accepted result, so you can compare this week's review against last quarter's.

Yes. Melaya's Device Control operates a real Android phone: opens the allowed app, reads the screen, taps and types, and pauses for on-device approval before anything is published. On this page the crew reads on-chain data through the blockchain, etherscan_tools, and mempool_tools bundles, so Device Control is only needed for balances that live inside an app with no public trail.

Yes, for the steps you route there. Melaya runs in the cloud or on your local runner, and per-step model routing across 23 providers lets you pin sensitive steps like the tax lot pass to local models. The blockchain bundle reads balances through your own RPC, and the knowledge store keeps your allocation policy and prior filings on the runner. Steps that call public explorers such as etherscan_tools do send addresses to that explorer, and the run trace shows exactly which steps did.

The etherscan_tools bundle surfaces every position in your transfer history, the core bundle fetches the protocol's audit pages, and dexscreener_tools checks whether exit liquidity is thinner than your entry was. The Report Writer then distills the findings into a one-page note with a single ACTION line and a confidence score, and pause_for_human holds any follow-up until you approve.