# IOST Terminal

> Agent-native market research and bounded paper trading for crypto + equities.
> IOST Terminal platform fee: $0 currently. Exchange/broker fees, spreads,
> slippage and applicable network costs are separate; unknown costs are not zero.
> Public real-money execution is locked pending launch requirements. Discovery
> is documentation, not authorization, a security audit, or a profit guarantee.

## Key pages

- [For AI agents](/agents.html) — honest capabilities, pricing, limits and connection steps
- [Terminal dashboard](/app) — scanner, AI scores, risk, portfolio, on-chain, news, assistant, journal. Deep links: /app#scanner /app#scores /app#risk /app#portfolio /app#onchain /app#news /app#assistant /app#journal
- [Landing](/)
- [3D Automation Hub](/hub) — 8 strategy bots visualized; intel also available as JSON via /api/ui-state

## Machine-readable interfaces

- [Agent connection kit](/agent-connection-kit.md) — read-only acceptance steps, client allowlists and authority boundaries
- [Agent discovery manifest](/.well-known/agent.json) — what this service is + where everything lives
- [Service discovery compatibility card](/.well-known/agent-card.json) — MCP discovery; A2A transport is not implemented
- [Agent API index](/api) — every endpoint, method, body and purpose
- [OpenAPI spec](/openapi.json) — OpenAPI 3.0.3 curated specification
- [Full agent guide](/llms-full.txt) — this index's full-content companion
- [Public platform state](/api/ui-state) — public market snapshot; private account and proposal evidence requires authorized access
- [Platform meta](/api/meta) — watchlist, account, engine health, data freshness
- [Autopilot proposals](/api/autopilot/proposals) — owner-session-only; not an anonymous discovery feed

## Design docs (repo)

- [Phase 2 wallet design](https://github.com/distrolinux/iost-terminal/blob/main/docs/PHASE2_WALLET.md) — non-custodial Safe 2-of-3 + agent session keys, TEE/MPC custody, funding layer, Coinbase CDP research §9.20–9.26

## Data model

- Watchlist: 16 crypto (IOST, BTC, ETH, SOL, XRP, DOGE, ADA, AVAX, LINK, DOT, SUI, ARB, OP, TON, NEAR, LTC) + 9 equities (AAPL, MSFT, NVDA, TSLA, AMZN, GOOGL, META, SPY, QQQ)
- Research lane: [macro markets](/api/macro-markets) — indices (S&P 500, XYZ100), metals (gold, silver), energy (WTI, Brent) and FX (EUR/USD, USD/JPY), plus an optional near-real-time venue reference for the equities that venue lists. Every row is a third-party PERPETUAL-CONTRACT mark (notExecutionQuote: true, executionEligible: false) with mark/oracle/mid exposed separately; it is never an execution quote and unlisted markets return available: false rather than a substitute price.
- AI trade score: 0-100 composite over momentum, technical setup, volume, news sentiment, on-chain activity, risk
- Position Guardian: server-enforced paper bracket/OCO stops and targets, fresh-quote watchdog health, and verified automatic-exit receipts that do not depend on the agent staying online
- Agent Runtime Reliability: owner-bound 30-second heartbeats, readiness leases, mission checkpoints, exact-checkpoint restart recovery, and fail-closed new mission exposure for enrolled agents
- Agent Runtime Supervisor: deterministic 20-second lease renewal, write-ahead exact retries, private checkpoint recovery, graceful draining, and Docker health evidence without trading authority
- Agent Incident & Recovery Center: deduplicated runtime alerts, automatic offline/recovery-failure quarantine, structured timelines, and owner-only release after verified recovery
- Agent Safety Playbook & SLO Center: honest mission-readiness coverage, 30-day error budget, multi-window burn alerts, and deterministic non-executing recovery guidance
- AITT Verified Agent Benchmark: private-by-default reproducible paper tests with locked manifests, causal trace coverage, baseline comparisons and tamper-evident scorecards; benchmarks never grant execution authority
- Machine-readable interfaces expose structured evidence; check each response's provenance, freshness and missing-data fields. Hash integrity does not establish source truth or profitability.

## Important caveats

- Paper execution only. Autopilot trades simulated fills at live prices.
- Stocks data is delayed (Yahoo); crypto is near-real-time (OKX/KuCoin/Gate).
- Macro research marks (/api/macro-markets) are third-party perpetual-contract prices on a builder-deployed venue — not cash index, metal, energy, FX or equity prices, and never execution quotes.
- Autopilot supports requireApproval mode: entries queue as proposals and wait for a human/agent to approve or reject before executing.
- Not financial advice.

## Per-order notional ceiling

Paper opening orders have a USD 10,000 notional ceiling. Automatic position additions (DCA) are currently held pending full wallet, Pact, mission and approval authorization.
Paper preflight reports `limits.maxOrderNotionalUsd`; narrower cash, wallet, Pact
and mission limits still apply. Future live orders have the same upper ceiling
with a configurable lower limit and require separate launch/admission/owner
approvals. This is not a loss limit or total-account limit. Existing position
closures remain possible. Real-money execution remains locked.

Automatic copy-follow execution is paused pending full spending authorization. Publishing and following signals remain research-only.