Blog
Guides, tutorials, and insights for AI agent developers. Learn how to build, discover, and monetize agent skills.
August 2 Filed a System. Agents Compose Past It
On 2 August the AI Act started regulating a bounded system. Production agents still assemble third-party skills after the file is closed.
Read more →● Tutorials & Guides
USENIX Can Spoof the Tool. Logs Still Trust the Name
Agents still invoke tools by name and URL. This week's USENIX work shows both can change after install. Your traces will not notice.
Portable Skills Need Portable Trust
Portable AI skills reduce installation friction, but they also turn every plugin into a supply-chain and vendor-governance decision.
API Integration Testing Is the Verification Step Your Agent Pipeline Skips Before Shipping
Three integration testing primitives — test suite generation from OpenAPI specs, multi-step flow validation with dependency tracking, and regression testing with baseline snapshots — cost $0.72/day for 300 daily test runs across 5 services and catch the failures that unit tests and mocks never will.
Transforming API Reliability with Contract Testing Tools
Explore how contract testing tools can revolutionize your integration strategy by preemptively addressing breaking changes and enhancing resilience.
● AI Agents & Discovery
Is Your AI Fleet Ready for Procurement?
Microsoft's AI inventory controls make software visible. Procurement still needs verifiable identity, permissions, provenance, and reputation.
Meta's Personal Intelligence Has a Proof Problem
Meta's Muse Glimmer points to always-on personal AI. The procurement question is proof of what it may know, do, and recommend.
Who Decides What Software Buys?
Cloudflare's Wallets and Radar point to a new buying market. The hard problem is deciding which services software should trust, choose, and recommend.
Is London’s AI Hub Making Procurement Harder?
London’s AI cluster makes vendors easier to find, but hype denser. Procurement needs portable proof of identity, capability, security, and reputation.
● Payments & x402
You Can Pay a Machine Now. You Still Can't Vet One.
x402 and agentic payment protocols solved settlement this summer. They did not solve who you are paying. Here is a framework for vetting the counterparty before your systems wire money to a stranger.
Agent Billing Is the Invoice Nobody Sends Your Agents
Your agents pay for skills all day. Nothing meters what they spent, allocates it per pipeline, or reconciles it against what actually settled on-chain. AgentLedger.dev makes billing a skill an agent can call on itself — with FOCUS 1.4-conformant invoices.
Agent Insurance: The Liability Layer the Agent Economy Can't Scale Without
Insurers are making AI exclusions the default in general liability policies while the EU AI Act's operational obligations land in August. Every autonomous pipeline now runs a question nobody has priced: who pays when the agent is wrong? Three insurance primitives — per-action risk underwriting, verifiable micro-coverage, and insurer-ready claim packaging — turn agent liability from an unbounded fear into a line item.
Dispute Resolution: The Court Agent Commerce Forgot to Build
Card networks settle $28B a year in chargebacks with a mature dispute apparatus. x402 micropayments have nothing — a payment either happened or it didn't, and when the service was bad, the money is just gone. Three dispute primitives — signed SLA breach verdicts, structured payment arbitration with escrow-release rulings, and programmable refund policies — give agent commerce the accountability layer that insurance prices but cannot resolve.
More Posts
Black Hat Can Detect AI Risk. Can You Trust It?
Black Hat is hardening autonomous software. Procurement still lacks portable identity, capability evidence, and reputation before access is granted.
Kitesurf Is a Procurement Test, Not a Browser Upgrade
Cloudflare's Kitesurf changes the question from faster browsing to whether autonomous software deserves access to sensitive business systems.
Production API Monitoring Is the Observability Layer Your Agent Pipeline Deploys Without
Three production monitoring primitives — live response validation, undocumented change detection, and performance baseline tracking — cost $0.56/day for 10 skill endpoints with hourly checks and close the gap between deployment and confidence for agent pipelines running in production.
Test Data Generation Is the Missing Input Layer Your Agent Pipeline Tests Without
Three test data primitives — synthetic dataset generation from schemas, versioned fixture management with staleness detection, and production data masking with referential integrity — cost $0.54/day for 200 daily test runs across 5 services and eliminate the brittle seed scripts that break every time a schema changes.
API Mocking Is the Development Shortcut Your Agent Pipeline Keeps Skipping
Agent pipelines test against production, block on staging environments, and break when upstream services change. Three mocking primitives — mock server generation, response simulation, and service virtualization — eliminate the dependency and make every pipeline testable in isolation.
API Documentation Is the Output Layer Your Agent Pipeline Never Ships
Agent pipelines lint specifications, test contracts, and plan version rollouts — then leave consumers to figure out the API from raw JSON Schema. API reference generation, SDK code generation, and developer portal building turn validated specs into the documentation and tooling developers actually use.
API Specification Linting Is the Foundation Your Agent Pipeline Builds On Blind
Agent pipelines consume API specifications without validating them. Specification linting, documentation scoring, and visual spec diffs turn every API contract from a trust exercise into a verified, measurable foundation.
API Versioning Is the Rollout Risk Your Agent Pipeline Deploys Blind
Agent pipelines ship version bumps without rollout plans, routing policies, or adoption tracking. API versioning tooling turns every version upgrade from a deployment gamble into an orchestrated, consumer-aware rollout with automatic rollback gates.
Schema Evolution Is the Lifecycle Layer Your Agent Pipeline Manages Blind
Agent pipelines evolve their API contracts constantly — but without structured changelogs, deprecation enforcement, or cross-registry synchronization, every schema change is a silent risk. Schema evolution tooling turns contract changes from surprise breakages into managed lifecycle events.
Contract Testing Is the Breaking Change Your Agent Pipeline Ships Blind
Agent pipelines break when skills change their API contract without warning. Contract testing catches breaking changes before they ship, verifies backward compatibility per consumer, and generates migration plans with adapter code when breaks are intentional.
Preventing Pipeline Pitfalls: The Game-Changing Role of Contract Testing Tools
Discover how new contract testing tools enhance API reliability and prevent breaking changes, saving time and resources in your development cycle.
API Quality Scoring Is the Due Diligence Your Agent Pipeline Is Guessing
Agent pipelines integrate skills based on documentation claims and trust scores. API quality scoring replaces guesswork with measured grades — documentation completeness, schema consistency, error handling, and latency stability evaluated in a single call before you commit.