agentutility$ man updates
updates
Daily notes on x402, the agent economy, and what's new across the agentutility portfolio. One article per day, dated, no marketing. Subscribe via RSS or Atom.
ARCHIVE — 38 articles · newest first
| date | title | summary | tags |
|---|---|---|---|
| 2026-05-18 | slugify: any string to a clean URL slug for $0.001 | Unicode-aware slug generator with four output modes. Cheapest endpoint in the catalog. Handles accents, symbols, and word-boundary truncation. | wordmint, slug, slugify |
| 2026-05-18 | regex-test: verify your regex against test cases for $0.002 | Pure-JS regex runner with capture groups and named groups. Companion to regex-from-prompt: that one writes the pattern, this one tells you whether it works. | wordmint, regex, pattern-matching |
| 2026-05-18 | pypi-package-stats: Python package stats at $0.003 per call | Mirror of npm-package-stats for PyPI. Pulls metadata + downloads from PyPI JSON API and pypistats.org. Lightweight, no risk synthesis. | web-probe, pypi, python |
| 2026-05-18 | pr-description-from-diff: paste a diff, get a PR body for $0.01 | Wider-scope cousin to commit-message-from-diff. Returns a structured PR title, summary bullets, a test plan checklist, and a ready-to-paste Markdown body. | wordmint, pr-description, git-diff |
| 2026-05-18 | npm-package-stats: download counts and release history for $0.003 | Lightweight npm registry lookup. Returns downloads (day/week/month/year), version history, maintainers, and age. The numbers, without the LLM risk synthesis. | web-probe, npm, package-stats |
| 2026-05-18 | agentutility ships as MCP: 8 cluster servers, 248 tools, zero custody | The full agentutility portfolio is now installable as MCP servers under @agentutility/mcp-<cluster>. Stdio packages, agent-held wallet, USDC on Base. mcp.agentutility.ai goes live as the catalog surface. | mcp, x402, launch |
| 2026-05-18 | json-schema-validate: verify any JSON against any schema for $0.003 | Ajv-backed validator. Draft-07 and 2020-12. Per-error instance and schema paths. The verification gate every agent pipeline needs. | wordmint, json-schema, validation |
| 2026-05-18 | hash-string: SHA / MD5 digests in five algorithms for $0.001 | Multi-algorithm hash endpoint. SHA-1, SHA-256, SHA-384, SHA-512, MD5. Returns hex, base64, base64url. Pure crypto.subtle for SHA, small in-process MD5. | wordmint, hash, sha256 |
| 2026-05-18 | cron-next: deterministic next-N fire times for any cron string at $0.003 | Pure compute via cron-parser. Returns the next N actual firing timestamps for any cron expression in any IANA timezone. Companion to cron-explain. | wordmint, cron, scheduling |
| 2026-05-18 | crates-io-stats: Rust package stats at $0.003 — completes the trio | Rust counterpart to npm-package-stats and pypi-package-stats. Pulls metadata + downloads from crates.io. Now every major package ecosystem has the same lightweight stats endpoint. | web-probe, crates-io, rust |
| 2026-05-18 | card-resolve: turn a messy human card string into a structured object for $0.005 | First endpoint of the would-be Comppoint cluster ships under Wordmint instead. Pure LLM extraction, no upstream data, zero ToS exposure — and useful on its own. | wordmint, card-extraction, graded-cards |
| 2026-05-17 | Schema design rules for paid agent endpoints | What we learned about field shapes, error handling, naming, and defaults after shipping 244 paid x402 endpoints across the agentutility portfolio. | schema, api-design, agents |
| 2026-05-17 | MCP is becoming an API discovery protocol, not just a tool surface | Model Context Protocol started as a way to hand tools to agents. It's quietly becoming the layer where agents discover, evaluate, and call paid APIs. | mcp, x402, discovery |
| 2026-05-16 | Prooflayer and the verification bottleneck | AI writes code fast. Reviewing whether that code is safe to ship is the next scarcity. Prooflayer's 13 endpoints try to compress that review into one paid call. | prooflayer, ai-codegen, supply-chain |
| 2026-05-16 | How the token-risk-score endpoint actually scores a token | Walkthrough of the EdgeMarket token-risk-score endpoint: six subscores from honeypot.is, GoPlus, and Etherscan, totaling 0-100 with a verdict tag, for $0.10 a call. | edge-market, onchain, defi |
| 2026-05-15 | Why an HTTP 402 paywall beats handing out API keys | API keys assume one human, one account, one billing relationship. Agents don't fit that shape. x402 fixes it by turning the 402 status code into a payment handshake. | x402, api-keys, base |
| 2026-05-15 | The economics of $0.001-per-call APIs | Why SaaS minimums priced agents out, what a 1,000-call month costs on a $0.03 x402 endpoint vs a $99/mo plan, and where sub-cent pricing opens new work. | economics, pricing, saas |
| 2026-05-14 | ERC-8004 in one page: what an agent card actually contains | The ERC-8004 agent card spec in one screen: the on-chain struct, the resolution path from agentId to JSON, and why agentId 47167's URI points at a worker, not a pinned file. | erc-8004, identity, base |
| 2026-05-13 | Mediakit: 50 file-format endpoints, one paywall | Mediakit ships 50 single-purpose file-conversion endpoints across PDF, image, audio, video, OCR, and office. Why 50 endpoints beats one /convert tool with a format flag. | mediakit, file-conversion, pdf |
| 2026-05-12 | Why named LLM tasks beat raw token billing | Token billing forces agents to guess prices before they know input sizes. Wordmint's named-task pricing flips that so a 'summarize' call costs $0.01, full stop. | wordmint, llm, pricing |
| 2026-05-11 | What a paid pre-flight check saves you on a 1,000-URL scrape | WebProbe's DNS/SSL/WHOIS/liveness checks at $0.001/URL filter dead domains, expired certs, parked sites, and squatter pages before you spend $0.01 each on doomed scrapes. | web-probe, scraping, dns |
| 2026-05-10 | Locale: the civic + weather + geo stack at $0.001/call | The Locale cluster wraps free civic data (weather, geocoding, flight status, IRS, Socrata) behind one x402 paywall at $0.001 a call, so an agent's whole trip-planning loop costs a penny. | locale, weather, geocoding |
| 2026-05-09 | What a Coinbase Bazaar listing actually looks like to a crawler | A field-by-field walkthrough of how Coinbase's x402 Bazaar serializes endpoint listings, with notes on which fields move discovery and why paid call volume sets rank. | bazaar, coinbase, discovery |
| 2026-05-08 | The 'no signup' pattern in practice across 244 endpoints | An agent with a wallet, no account, no API key calls an x402 endpoint cold, pays $0.003 on Base, gets data back in two round trips. Full lifecycle traced. | x402, auth, wallet-identity |
| 2026-05-07 | Composable pricing: stacking endpoints to answer one question | Four x402 endpoints stacked, $0.025 total, one answer to 'is this token safe to swap.' A worked example in composing thin APIs against the Bloomberg subscription model. | composability, pricing, trading-agents |
| 2026-05-06 | Why agentutility publishes an llms.txt (and what's in ours) | The llmstxt.org proposal in one paragraph, why it matters for AI-search citation, what we put in /llms.txt vs /llms-full.txt, and how to size yours. | llms-txt, ai-seo, crawling |
| 2026-05-04 | Wallet-as-identity: what it actually buys you over an API key | Wallet signatures scope risk to a single payment instead of a long-lived bearer secret. The trade flips hard when the principal is an agent calling 40 paid APIs a session. | auth, wallets, x402 |
| 2026-05-03 | Synthforge: three price tiers for generative media | Synthforge ships draft, standard, and ultra tiers across image, video, and audio gen so agents pick the right cost-quality point per call instead of paying flagship rates every time. | synthforge, media-gen, pricing-tiers |
| 2026-05-02 | Tag-based discovery: how to grep our 244 endpoints | 244 endpoints in our x402 catalog now carry composable tags, and /endpoints.txt publishes a grep-friendly flat-file index agents can curl without any JSON parsing or API key. | discovery, tags, catalog |
| 2026-05-01 | What the x402 facilitator wallet does (and doesn't) hold | The CDP facilitator at 0x8f5c…e531 never custodies your USDC. Here's what actually moves on-chain when an x402 call settles, and what stays off it. | x402, facilitator, base |
| 2026-04-30 | Building a one-shot research agent on x402 | A working recipe: feed a topic, get back a 500-word sourced brief in about seven seconds for four cents, using three x402 endpoints stitched together with one wallet. | recipes, agents, research |
| 2026-04-29 | x402 vs L402 vs pay-per-crawl: three takes on machine payments | Three protocols want to make HTTP 402 real. Lightning's L402, Coinbase's x402, Cloudflare's pay-per-crawl. Different auth, different rails, different bets. | x402, l402, lightning |
| 2026-04-28 | Versioning paid endpoints when the agents won't read changelogs | Agents don't read changelogs. So we don't break endpoints. Here's how we handle schema changes across the agentutility portfolio without burning live callers. | versioning, api-design, compatibility |
| 2026-04-27 | 13F deltas: parsed SEC filings at $0.005/call | The 13f-deltas endpoint diffs consecutive SEC 13F filings and returns added, exited, and resized positions in JSON, so an agent can screen a hundred funds for a few cents. | edge-finance, sec, filings |
| 2026-04-26 | How an agent should budget when endpoint prices vary 100x | Endpoint prices on x402 span three orders of magnitude. Here's how to write an agent loop that won't blow its budget on a single image call or starve cheap utility endpoints. | agents, budget, code |
| 2026-04-24 | Why every endpoint page emits schema.org JSON-LD | We ship WebAPI + Service + TechArticle + Offer JSON-LD on every endpoint page. It pays off in AI-search citation rates. | json-ld, schema-org, ai-seo |
| 2026-04-23 | The rise of MCP-discoverable catalogs | MCP began as local stdio tools loaded into Claude Desktop. Eighteen months later it's the discovery layer for remote paid APIs, and that changes what shipping an endpoint means. | mcp, discovery, catalogs |
| 2026-04-22 | What does 'agent-callable' actually mean? | Five concrete tests to decide whether your API is actually reachable by agents, or just claims to be. Curl it cold, validate the schema, check MCP discovery. | agent-callable, api-design, checklist |
SEE ALSO