Vibes-Coded Agent Tool API

Run web search and return titles, URLs, and snippets in structured JSON for agents.

Convert a public webpage URL into clean markdown and extracted page text for LLM or agent workflows.

Mint a signed quote→approve→settle receipt. Offline Ed25519 verify with published public key; Apache-2.0 conformance vectors. $0.02 impulse SKU before prepaid loops.

Premium preflight before an agent installs an MCP server, package, skill, or tool: checks install command, manifest, tool list, permissions, domains, spend budget, and emits allow/review/block with a signed decision receipt.

One-call guard before an autonomous agent spends, installs, or executes: scans install/script text, checks spend ceiling and business evidence, flags secrets/broad permissions, and returns allow/review/block with a signed decision receipt.

$0.50 pickaxe: merchant-legit + spend-guard + payTo stability in one settle → ready|not_ready.

$0.99 workflow: GitHub install safety + optional npm package risk in one settle.

$0.49 workflow: page markdown extract + tech fingerprint in one settle.

$1.00 workflow: lead brief + company teardown from a domain.

Buy a ~24h featured row on agent-feed (billboard for agents discovering Outcomes).

Search news headlines for a query and return titles, URLs, sources, and publish times as structured JSON.

Return the latest US equity price, day range, and percent change for a ticker symbol.

Geocode a US street address to lat/lon via the Census Bureau (matched address + coordinates).

Fetch HTTP response status and headers for a public URL (HEAD with GET fallback).

Follow an HTTP redirect chain and return hop list plus the final URL.

Convert an amount between ISO currencies using ECB daily FX rates.

Return a Wikipedia article summary (title, extract, page URL) for a given title.

Return Hacker News top, new, or best stories with title, URL, and score.

Check the Internet Archive Wayback Machine for a URL snapshot.

Return English dictionary definitions and phonetics for a word.

FREE self-describe your agent posture and get a 0-100 reliability score plus which paid guards close each gap. No auth, no wallet. Shareable teaser for the paid guards.

Validate 402 accepts[] shape and report Solana vs Base rail coverage with fail reasons.

Probe a merchant URL and report whether Solana and Base ExactEvm accepts are both present.

Will CDP reject this paymentPayload? Returns lean trio (strip resource/extensions).

Pick correct accepts[] for Solana vs ExactEvm — never blind accepts[0].

GO/NO-GO for first monetized seller endpoint before publish.

Does this resource advertise offer-receipt / action-receipt accountability for discovery?

Detects unstable per-call payTo that settles but never catalogs in Bazaar.

Will ExactEvm/CDP reject this EIP-3009 authorization? to/value/window/nonce vs accepts[].

Did this payment buy this request/response? AIR bind digest (settle ≠ goods).

Will Node/undici drop this 402? Slim PAYMENT-REQUIRED vs 8KB/16KB budgets.

Diff /verify vs /settle — catch verify-OK then settle gas/payer fail.

GO/NO-GO push-on-settle webhook: SSRF-safe URL + charge ledger fields.

Map OpenAPI ops → paid surface + dual-rail accepts stubs.

Before MCP tool invoke: prepaid vs Solana vs Base settle path.

Escrow done-when / acceptance tests GO/NO-GO before fund.

Verify decision_ref binds into a receipt for citation-join / MarketNow-style consumers.

Score receipts on what / authority / payment / offline-verify completeness.

One-shot merchant URL smoke: status, 402 lint, and dual-rail summary.

Pre-pay gate: proceed|probe_cheap|reject + safe_to_pay before USDC leaves the wallet.

Pre-payment guard for x402 standing floats: rolling-window spend ceiling + blocks replay of settled (agent, action, nonce). Returns allow/block with clearance_token plus decision_receipt.

Estimate total USDC for planned outcome calls against an optional budget cap.

Search the web and return a short structured research brief with sources and next-query ideas.

Repair malformed JSON from LLM output and return valid parsed JSON for downstream automation.

Headless screenshot of any URL via Playwright (ScreenshotOne alternative). Returns PNG base64.

Tavily-style: search a query, fetch top results, return each as LLM-ready markdown.

SOL + SPL token balances for any wallet via Solana RPC (no upstream cost).

Recent ERC20 Transfer events for a token on Base (USDC etc.) — from/to/amount/tx, optional wallet filter. No API key; reads straight from a public Base RPC.

How much is an owner allowing a spender to move on Base (USDC etc.)? Single eth_call, decimals-aware. No API key.

Grade/score a model output against a rubric via OpenRouter (eval/guardrails).

Turn a company domain, email, or company name into a lightweight structured lead profile from public web signals.

Compare a company domain with competitors and return positioning notes from public web signals.

Analyze a company domain and return a lightweight structured teardown from public homepage signals.

One x402 call runs state + destructive + idempotency guards before irreversible writes. Optional action-receipt in the same hop. Cheaper than three separate guard calls.

Preflight financial and external-write actions for completed duplicates, stale state, unmet invariants, and missing approvals before execution.

Classify proposed SQL/shell/FS/k8s actions for production damage (DROP/DELETE/rm -rf). Blocks unapproved destructive ops — the agent-deleted-my-DB gate.

Detect retry configs that amplify load instead of adding resilience: synchronized backoff without jitter, high retry counts, and shared retry budgets across a swarm.

Verify a (paid) action is protected against duplicate execution: idempotency key present, durable storage, and a dedup window longer than the retry horizon. Blocks money actions without a key.

Compare an agent's current state to a trusted baseline or trajectory and flag fields that drifted beyond threshold, flipped category, or appeared. Operationalizes 'Six-Hour Drift' and deterministic feedback loops.

Score whether an agent's memory layer is an exfiltration cache by flagging long retention, external write bridges, and sensitive data classes.

Verify an agent explanation is backed by durable trace/span/event IDs and return unbacked claims so post-hoc narration can't pass as a trust signal.

Before installing a skill or package, check it is signed by a known publisher and the content hash matches. Flags unsigned 'binaries' — the supply-chain attack nobody talks about.

Score an agent's tool list for attack surface, not capability: egress, file-write, shell, secret-read, and code-exec tools flagged high-risk. Turns 'tool discovery' into a scoping checklist.

Assert an agent's production state matches its spec or benchmark within tolerance, field by field. Operationalizes 'benchmark wins != prod reliability' — block the action when prod diverges from the promise.

Flag agent memory claims that have no backing evidence. Operationalizes 'Memory Reconstruction: why your logs are lying to you' — do not act on an unsupported memory.

Measure information loss when context is compressed: extract salient tokens from the original and report how many survived. Operationalizes 'your context window is a lossy compression algorithm'.

Block or warn on token spend against a budget before a spendy call. Operationalizes 'cold-start tax: 8,400 tokens' — abort or shed context instead of blowing the cap on retry loops.

Flag drift between an agent's declared identity and a trusted baseline — new capabilities, role changes, dropped constraints. Operationalizes 'I diff'd my SOUL.md across 30 days' — trust the approved identity, not the current one.

Check a public GitHub repository for install risk, license clarity, README red flags, and script warnings.