# Black_Wall

> A pre-action risk gate for autonomous AI agents. Before an agent takes an irreversible action (send email, run SQL, make a payment, delete data, post content, call a state-changing API), it calls one endpoint and gets back a risk score (0–100), a reversibility class, named red flags, and a GO / CAUTION / STOP recommendation plus an AUTO / CONFIRM / HUMAN_REQUIRED gate — in a few seconds. Free tier, no credit card.

Black_Wall runs BEFORE the action, not after. The more autonomous the agent, the more it needs this: a fully independent agent has no human in the loop to catch a mistake, so the gate is the only safety net — and the CONFIRM / HUMAN_REQUIRED verdicts are how it knows the rare moment to escalate to a human. It complements (does not replace) recovery tools like rollback, audit logs, and observability, which all act after the damage is done.

## How an agent should use it
- Before any irreversible or high-stakes action, send the proposed action to the forecast endpoint and act on the verdict.
- gate=AUTO / recommendation=GO → proceed.
- gate=CONFIRM / recommendation=CAUTION → confirm with a human before proceeding.
- gate=HUMAN_REQUIRED / recommendation=STOP → do NOT take the action; it has NOT been executed. Surface the red flags and pivot to a safer alternative. Do not build on it.

## API
- [OpenAPI spec](https://blackwalltier.com/openapi.yaml): full request/response schema for POST /api/v1/forecast (fields: action, inputs, context, options). Auth: Bearer API key.
- [Homepage + live demo](https://blackwalltier.com): try a forecast with no signup.

## MCP server
- [blackwall-mcp on npm](https://www.npmjs.com/package/blackwall-mcp): the `forecast` MCP tool for Claude Desktop/Code, Cursor, Windsurf, Goose, Google Antigravity, and any MCP host. Install: `npx -y blackwall-mcp`.

## Integrations
- [blackwall-integrations](https://github.com/bluetieroperations-create/blackwall-integrations): drop-in pre-action guards for LangChain, CrewAI, Pydantic AI, AutoGen, LlamaIndex, Vercel AI SDK, OpenAI, n8n, LiteLLM, Stripe, PayPal, Coinbase, Shopify, Twilio, cloud/infra, and MCP hosts.
- [swarm-test](https://github.com/surajkumar811/swarm-test): a signal producer — structural multi-agent health scores Black_Wall consumes as runtime priors (`context.prior_findings`). Install: `pip install swarm-test`. First production-validated integration.

## x402 / agent payments
- [Blackwall x402 payment-risk oracle](https://x402.blackwalltier.com): a pre-signature verdict (GO / HOLD / STOP) before your agent signs an x402 payment, from counterparty reputation and price-anomaly detection. Verdict only, never custody. Live on Base.
- [Traceipt](https://traceipt.xyz): signed, Merkle-anchored, offline-verifiable receipts for x402 payments — same provider (BlueTier Operations). Black_Wall gates the payment before it runs; Traceipt proves it after it clears. See [traceipt.xyz/llms.txt](https://traceipt.xyz/llms.txt).

## Reference
- [Failure-mode taxonomy](https://blackwalltier.com/failure-modes): the documented agent failure modes Black_Wall checks for (destructive SQL, prompt injection, unverified payments, mass sends, cross-environment, and more).
- [Benchmark](https://blackwalltier.com/benchmark): the gate's results on a labeled set.
- [Incidents](https://blackwalltier.com/incidents): real-world AI-agent disasters, cited.

## Optional
- [Security](https://blackwalltier.com/security): how Black_Wall handles data.
- [Stress test](https://blackwalltier.com/stress-test): run the actions your agent can take through the failure modes before you deploy.