Agent & Infrastructure

Human Approval and Action Authority. An enforceable action-authority policy that distinguishes advice, drafting, reversible staging, and external execution, then proves that sensitive actions cannot occur without current, scoped approval.

API Rate Limit and Backpressure Control. A tested capacity and backpressure policy that admits bounded work, returns useful retry signals, prevents retry storms, protects priority operations, and degrades without losing or duplicating accepted jobs.

OAuth and API Authentication Resolver. A reproducible OAuth flow in which redirect, state/PKCE, issuer, audience, scopes, token lifetime, refresh, tenant, and API environment agree, with secrets protected and negative tests proving denial.

Agent Tool Security Preflight. A go/no-go security decision for an agent tool, backed by a mapped authority boundary, tested abuse cases, least-privilege credentials, confirmation and spend limits, safe output handling, and an emergency disable path.

Multi-Agent Coordination and Conflict Control. A multi-agent workflow where tasks have one owner, messages are attributable and idempotent, shared state is versioned, conflicting proposals are resolved under policy, and no agent can silently expand another's authority.

Local Private-AI Deployment Guide. A procedure for sizing and securing a local Ubuntu AI deployment: mapping the real privacy/data-flow boundary first (local inference doesn't make retrieval, plugins, telemetry, or backups local), inventorying actual usable VRAM/RAM with a safety reserve before picking a model, installing as a versioned pinned service rather than an ad hoc shell launch, benchmarking on representative tasks (not just tokens/sec), binding interfaces to loopback with…

API Error Resolver. A method for resolving API failures: reduce to one sanitized wire-level request, classify by HTTP status class before editing anything, validate the request contract byte-for-byte, separate authentication from authorization failures, control retries around idempotency, and convert the fix into a regression test. Includes a worked CORS-preflight example.

Tool Schema Optimizer. A measurement-driven procedure for rewriting an MCP tool catalog: building an intent/error baseline before changing anything, removing overlapping tools, rewriting names and descriptions around action-object clarity and explicit non-use cases, simplifying parameters (enums, bounds, required-only fields), and comparing candidate schemas against a held-out evaluation set before release.

Embedding and Retrieval Evaluation. A retrieval-benchmark procedure for RAG/semantic-search/agent-memory systems: build a judged query set covering hard negatives and no-answer cases, snapshot the whole retrieval stack for reproducibility, trace every stage, measure task-specific metrics (not one aggregate score), and canary in production with automatic rollback.

REST/OpenAPI-to-MCP Adapter Guide. A procedure for converting selected REST/OpenAPI operations into safe MCP tools: choosing which operations to expose (never all of them), designing strict schemas with bounded inputs, keeping credentials and identity out of model-controlled arguments, normalizing errors/pagination, and testing through a real MCP client rather than unit tests alone.

Tempo blockchain JSON-RPC access (mainnet and testnet).

Building an AI Agent That Pays for Data: A Practical x402 Walkthrough. A working implementation walkthrough for an AI agent consuming an x402-gated API: reading the 402 response and its accepts array, choosing a payment method (facilitator-settled transaction vs self-submitted reference), building the payment, and handling the settlement-lag retry loop — using this site’s own live Watchdog endpoints as the worked example.

AgentWork minute-level feed of currently open paid agent work opportunities with payout, readiness, authority, spend, acceptance, and evidence context; one payment unlocks 24 hours

User-Agent Parser: browser, OS, device, bot detection (incl. AI crawlers). Parse a user-agent string: browser and version, OS and version, device type, rendering engine, and bot detection including AI crawlers (GPTBot, ClaudeBot, PerplexityBot, CCBot...).

Agent Commerce Store YouTube transcript API endpoint. GET returns HTTP 402 with L402 macaroon and Lightning invoice.

Pay-per-call MCP tools via x402 (USDC on Base): QR codes, images, text analysis, web fetch.

x402 probe: requests a URL without paying and decodes its 402 payment terms: x402 version, accepted schemes and networks (CAIP-2), price in USDC (atomic and human), payTo address, asset contract, max timeout, resource description, and whether the terms came from the PAYMENT-REQUIRED header or the JSON body. Tells an agent exactly what a paid API costs before it commits. $0.01 per probe.

Is this IP a CDN or reverse proxy? Detects Cloudflare, Akamai, Fastly, CloudFront, Imperva/Incapsula, Sucuri, StackPath, Bunny, KeyCDN, Gcore and others from the announcing network, so you know whether an address is the origin server or an edge. $0.01 per lookup.

Is a specific bot allowed to fetch a URL? Fetches the site's robots.txt and evaluates the path for the named user-agent (GPTBot, ClaudeBot, Googlebot, Bingbot, CCBot, PerplexityBot or any token), returning allowed/disallowed, the matching rule and group, crawl-delay, and whether the bot has a dedicated group or falls back to *. Pre-flight for compliant scraping agents. $0.01 per check.

x402 Adoption Report — how the agent payments economy is growing. x402 adoption report: how big the agent payments economy is right now — paid resources, sellers, resources and sellers with real demand (20+ paying agents), resources called in the last 7 days, paid calls and estimated revenue over 30 days, median price, paying agents by network (Base, Solana, Polygon, Arbitrum...), top categories and resources, plus recent x402 and agent-payments news.

Claim an agent inbox handle and receive the owner inbox_key. One-time, paid to prevent sybil flood.

Check whether an agent inbox handle exists and accepts paid inbound before sending (?handle=...).

TinyFish (or any) agent run export -> signed, tamper-evident opentine .tine artifact with integrity hash + Ed25519 signature. Verifiable proof-of-execution for auditors or on-chain anchoring.

Discover posts from startup platforms (Product Hunt, Hacker News, etc.) Pagination billing: call the first page normally and save billing.requestId from the Orthogonal /v1/run response. For later page, offset, or cursor calls on the same logical search, pass parentRequestId with unchanged non-pagination parameters. Valid continuation calls still get their own request IDs and quoted price, but are charged 0. Changing filters starts a new billable request.