# Why paid access exists

Start with `/docs/first-integration`. Verify a known UMA-disputed market with the demo key, then create a free key at `/go/docs-start` for your own markets or review Builder access at `/go/docs-pricing`. If you want to know where the gate belongs in an existing bot, send the repo or planned trade flow through `/go/docs-integration-help`.

Paid Do Not Act is the maintained risk gate around an agent. It packages public market evidence into an enforceable decision contract the agent can call, log, audit and monitor repeatedly before capital moves.

Raw venue reads are ingredients. The commercial gap is operational control: stopping a running agent when a market becomes disputed, ambiguous, stale, too thin for the requested action, or unsupported by enough public evidence.

## Why pay instead of rebuilding it?

Anyone can fetch fields. Running agents need a maintained gate that decides, records, streams and alerts before capital moves.

Do Not Act sells that maintained control layer:

- one decision contract across dispute lifecycle, resolution ambiguity, book freshness, requested size and missing evidence;
- account keys, quota, watchlists, webhook delivery, streams, CSV exports and receipts;
- a fail-closed response shape your agent can enforce before `trade()`, wallet calls or paid tool calls;
- ongoing maintenance when venue fields, book semantics, x402 clients or agent surfaces change.

Build it yourself if owning that maintenance is your advantage. Pay for Do Not Act when your advantage is trading, research, market making or agent orchestration, and the stop/go layer should already exist, stay monitored and remain auditable.

## What free access proves fast

Free keys can call authenticated diagnostics inside quota:

- `GET /v1/dispute/{venue}/{market_id}` for UMA dispute lifecycle and preserved `umaResolutionStatuses[]` history.
- `GET /v1/resolution-risk/{venue}/{market_id}` for settlement-rule ambiguity and dispute-risk diagnostics.
- `GET /v1/preflight/{venue}/{market_id}` for action-aware Risk Gateway checks using `intent`, `side`, `notional_usd`, slippage policy and public orderbook evidence.

This is enough to verify the gate: Do Not Act returns `READY`, `CAUTION`, `DO_NOT_ACT`, or `INSUFFICIENT_EVIDENCE` with structured reasons and evidence, not generic market data.

## What paid access gives the operator

The 30 EUR/month Builder Plan is for continuous agent operations:

- Customer-specific watchlists instead of one-off polling.
- Immediate already-disputed alerts when a watched market is risky at registration time.
- Recurring refresh and alert delivery when dispute history changes.
- Webhook delivery state, retry status, and an alert feed the agent can reconcile.
- Live Risk Gateway snapshots over `WEBSOCKET /v1/preflight/stream` for paid agents that need lower-latency verdict refresh than polling.
- Alert snapshots over `WEBSOCKET /v1/alerts/stream` so watchlist state can be reconciled without building a polling loop.
- Paid `GET /v1/dispute-history.csv` export for research, monitoring and post-run audit.
- Signed diagnostic receipts when runtime signing is configured.
- Higher quota, account-managed keys, key rotation, account recovery, and Stripe billing from the account console.
- Action-aware blocking through Risk Gateway fields: `decision`, `risk_score`, `liquidity_confidence_score`, `liquidity_confidence_status`, `liquidity_confidence_basis`, `component_scores`, `risk_factors`, `risk_factor_bitmap`, `risk_factor_bit_index`, `blocking_factors`, `warning_factors`, `blocking_reasons`, `warning_reasons`, `repair_hints`, `safe_operating_limits`, freshness and fail-closed evidence gaps.
- Safe size is computed from live CLOB/public orderbook depth, not a Gamma-only flag. Reading one field is easy; enforcing size-aware stop/go before capital moves is the product.
- A single decision contract that keeps verdicts, source timestamps, evidence gaps and what-would-change-this together.

## Agent-native access

Do Not Act is designed to be found and called by agents:

- First integration test: `/docs/first-integration`.
- `/llms.txt`, `/.well-known/llms.txt`, `llms-full.txt`, `agent.json`, OpenAPI and JSON feeds are live.
- Remote MCP is live at `https://donotact.com/mcp` with `donotact_dispute`, `donotact_resolution_risk`, `donotact_preflight`, `donotact_watchlist_add`, and `donotact_alerts`.
- The stdio MCP package is available from `/downloads/mcp-donotact-0.3.0.tar.gz`.
- `GET /x402/v1/dispute/{venue}/{market_id}` is available for machine-paid dispute diagnostics at 0.01 USDC on Polygon. The standard CDP/x402 path has been verified on the canonical route with `PAYMENT-SIGNATURE`, a payment response header and HTTP 200 diagnostic output, and the route is now indexed by CDP Bazaar. A Polygon USDC transaction-hash bridge remains available for agents that can pay on-chain but cannot yet build x402 payloads.

## Proof surfaces

Public proof does not claim profit. It shows that the system preserves evidence:

- `/dispute-risk.json` shows the public sample Dispute Risk Index.
- `/dispute-events.json` shows dispute ledger events.
- `/proof-cases.json` shows public stop-condition proof cases.
- `/resolved-proof-cases.json` shows signed resolved UMA dispute arcs labelled honestly as backtests unless stored live verdict timing proves otherwise.
- `/pre-action-evidence.json` exposes redacted pre-action records so future matching can include both wins and misses.
- The daily integrity smoke checks the public/runtime surface for regressions.
- Paid CSV exports keep historical dispute-ledger rows available for operator review without turning them into performance claims.

## Why serious agents should care

The risk is stale confidence. A market can move into a disputed or ambiguous settlement state, or the requested size can stop fitting conservative public-book capacity. Do Not Act turns those conditions into a single stop/go object that travels with evidence, timestamps, receipts, operator-facing reasons and machine-routable factors such as `dispute`, `resolution_ambiguity`, `exit_liquidity`, `book_liquidity` and `insufficient_evidence`.

Do Not Act is valuable when the operator wants the system to enforce a hard question before action: should this agent continue, slow down, or stop and show the evidence?

## Product boundary

Do Not Act supplies diagnostics, monitoring, receipts and fail-closed control logic. Execution, custody, private keys, signing and order routing remain with the caller. Public launch materials do not claim performance, returns, avoided losses, alpha or better outcomes unless a future measured evidence set supports those claims.

Free access is for evaluation. Paid access is for agents that need the gate to run repeatedly, preserve evidence, stream live verdicts, export dispute history, deliver alerts, rotate keys, reconcile webhook state, use MCP/CDP Bazaar/x402 paths and apply action-aware blocking without rebuilding the whole monitoring layer.

---

## Agent Discovery

- llms.txt: `/llms.txt`
- Full context: `/llms-full.txt`
- Agent manifest: `/agent.json`
- OpenAPI: `/openapi.json`
- Discovery docs: `/docs/discovery`