<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>airoweb</title><description>Notes on AI workflows, tools, and MCPs.</description><link>https://airoweb.com/</link><item><title>Agent Browser is a browser operator, not a prompt pack</title><link>https://airoweb.com/blog/agent-browser-operator-boundary/</link><guid isPermaLink="true">https://airoweb.com/blog/agent-browser-operator-boundary/</guid><description>Vercel Labs&apos; Agent Browser skill can turn a coding agent into a web operator. Adopt it with a browser, account, and permission boundary of its own.</description><pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate></item><item><title>An AI renewal is a workflow review, not a seat-count negotiation</title><link>https://airoweb.com/blog/ai-tool-renewal-workflow-review/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-tool-renewal-workflow-review/</guid><description>A decision memo for reviewing AI tools against the work they perform, the risks they introduced, and the exit options the organization still controls.</description><pubDate>Sun, 20 Sep 2026 00:00:00 GMT</pubDate></item><item><title>Jev belongs in the decision layer, not the writing layer</title><link>https://airoweb.com/blog/jev-decision-layer-use-cases/</link><guid isPermaLink="true">https://airoweb.com/blog/jev-decision-layer-use-cases/</guid><description>Three practical Jev use cases for code review, strategy portfolios, and marketing operations—and the controls each workflow still needs.</description><pubDate>Sun, 20 Sep 2026 00:00:00 GMT</pubDate></item><item><title>NotchFuel puts AI usage in the MacBook notch</title><link>https://airoweb.com/blog/notchfuel-ai-usage-in-macbook-notch/</link><guid isPermaLink="true">https://airoweb.com/blog/notchfuel-ai-usage-in-macbook-notch/</guid><description>A small open-source macOS app turns the camera notch into an expandable usage gauge for Claude Code, Codex, and Grok CLI.</description><pubDate>Mon, 31 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Put the AI pentester in a test cell</title><link>https://airoweb.com/blog/ai-penetration-testing-agent-guardrails/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-penetration-testing-agent-guardrails/</guid><description>A controlled operating model for using AI agents in SaaS penetration tests without turning a useful assistant into an unbounded attack tool.</description><pubDate>Tue, 25 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Ask workers where the AI moved the work</title><link>https://airoweb.com/blog/ask-workers-where-ai-moved-work/</link><guid isPermaLink="true">https://airoweb.com/blog/ask-workers-where-ai-moved-work/</guid><description>AI adoption often shifts review, exceptions, and accountability onto frontline teams. Find that hidden work before the rollout hardens around it.</description><pubDate>Sun, 16 Aug 2026 00:00:00 GMT</pubDate></item><item><title>The agent updated the record. The notification never left.</title><link>https://airoweb.com/blog/agent-transactional-outbox/</link><guid isPermaLink="true">https://airoweb.com/blog/agent-transactional-outbox/</guid><description>A transactional outbox keeps an agent&apos;s database decision and its downstream side effect from drifting apart when a process crashes or a network call fails.</description><pubDate>Mon, 03 Aug 2026 00:00:00 GMT</pubDate></item><item><title>AI governance needs decision rights, not another committee</title><link>https://airoweb.com/blog/ai-governance-decision-rights/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-governance-decision-rights/</guid><description>A practical operating model for separating central AI governance from ownership of the business workflows that use AI.</description><pubDate>Sun, 02 Aug 2026 00:00:00 GMT</pubDate></item><item><title>A running flag is not a lock for your agent queue</title><link>https://airoweb.com/blog/agent-queue-leases/</link><guid isPermaLink="true">https://airoweb.com/blog/agent-queue-leases/</guid><description>A practical design for claiming agent jobs with expiring leases, guarded heartbeats, and fencing that stops stale workers from writing.</description><pubDate>Mon, 27 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Stop measuring AI literacy by course completion</title><link>https://airoweb.com/blog/stop-measuring-ai-literacy-by-course-completion/</link><guid isPermaLink="true">https://airoweb.com/blog/stop-measuring-ai-literacy-by-course-completion/</guid><description>A role-based operating model for showing that people can use and supervise AI in their actual work, without turning training records into compliance theater.</description><pubDate>Sun, 26 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Valid JSON is not permission to act</title><link>https://airoweb.com/blog/valid-json-is-not-permission-to-act/</link><guid isPermaLink="true">https://airoweb.com/blog/valid-json-is-not-permission-to-act/</guid><description>Structured model output solves a parsing problem. Production agent workflows still need business-rule, authorization, and execution checks before a write.</description><pubDate>Mon, 20 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Measure the AI workflow, not how often people open the tool</title><link>https://airoweb.com/blog/measure-ai-workflow-not-tool-usage/</link><guid isPermaLink="true">https://airoweb.com/blog/measure-ai-workflow-not-tool-usage/</guid><description>A practical scorecard for judging AI workflow outcomes, rework, risk, and operating cost without mistaking product activity for business value.</description><pubDate>Sun, 19 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Give your agent a key, not the keyring</title><link>https://airoweb.com/blog/scope-agent-credentials-to-the-run/</link><guid isPermaLink="true">https://airoweb.com/blog/scope-agent-credentials-to-the-run/</guid><description>How to provision, scope, and expire the credentials an AI agent uses so a single run cannot borrow the whole company&apos;s access.</description><pubDate>Mon, 13 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Design the AI workflow&apos;s off switch before you need it</title><link>https://airoweb.com/blog/design-the-ai-workflow-off-switch/</link><guid isPermaLink="true">https://airoweb.com/blog/design-the-ai-workflow-off-switch/</guid><description>Teams review AI intake and pilots but never plan the shutdown. Here is the operating model for retiring an AI workflow on schedule and killing one in a hurry.</description><pubDate>Sun, 12 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Assume your agent will run the workflow twice</title><link>https://airoweb.com/blog/assume-agent-runs-workflow-twice/</link><guid isPermaLink="true">https://airoweb.com/blog/assume-agent-runs-workflow-twice/</guid><description>Agent workflows retry, time out, and re-fire background steps. Here is how to make the acting steps safe to repeat with idempotency keys and dedup.</description><pubDate>Fri, 10 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Your agent logs are a mirror: a self-audit from AI session history</title><link>https://airoweb.com/blog/agent-logs-are-a-mirror/</link><guid isPermaLink="true">https://airoweb.com/blog/agent-logs-are-a-mirror/</guid><description>Your AI session history is a behavioral record you never posed for. Here is how to mine it into an honest work audit, run it safely, and what it gets wrong.</description><pubDate>Wed, 08 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Give coding agents a workbench, not a workstation</title><link>https://airoweb.com/blog/coding-agent-workbench-boundary/</link><guid isPermaLink="true">https://airoweb.com/blog/coding-agent-workbench-boundary/</guid><description>How engineering teams should bound coding-agent environments before granting shells, network access, secrets, or repository write paths.</description><pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Do not send every AI request through the same approval queue</title><link>https://airoweb.com/blog/ai-governance-intake-lanes/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-governance-intake-lanes/</guid><description>A decision memo for separating AI requests into local, reviewed, and restricted intake lanes before governance becomes a bottleneck.</description><pubDate>Sun, 05 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Agent workflows need an evidence trail before they need more autonomy</title><link>https://airoweb.com/blog/agent-workflow-evidence-trail/</link><guid isPermaLink="true">https://airoweb.com/blog/agent-workflow-evidence-trail/</guid><description>A practical operating model for logging tool calls, approvals, sources, and handoffs in connected AI agent workflows.</description><pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Review AI workflows again when the workflow changes</title><link>https://airoweb.com/blog/ai-workflow-change-review/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-workflow-change-review/</guid><description>A decision memo for deciding when an approved AI workflow needs a fresh review instead of waiting for the next scheduled audit.</description><pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Keep SOUL.md out of the agent junk drawer</title><link>https://airoweb.com/blog/keep-soul-md-out-of-the-junk-drawer/</link><guid isPermaLink="true">https://airoweb.com/blog/keep-soul-md-out-of-the-junk-drawer/</guid><description>A practical guide to keeping an agent SOUL.md focused on durable identity while moving project rules, workflows, and tool settings elsewhere.</description><pubDate>Thu, 02 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Treat MCP servers like production APIs</title><link>https://airoweb.com/blog/mcp-servers-production-apis/</link><guid isPermaLink="true">https://airoweb.com/blog/mcp-servers-production-apis/</guid><description>A practical operating model for teams moving MCP servers from local experiments into governed production infrastructure.</description><pubDate>Thu, 02 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Treat Multica skills as operating knowledge, not prompt storage</title><link>https://airoweb.com/blog/multica-skills-operating-knowledge/</link><guid isPermaLink="true">https://airoweb.com/blog/multica-skills-operating-knowledge/</guid><description>A decision memo for teams deciding when a Multica skill should become shared agent capability, and when it should stay local or manual.</description><pubDate>Thu, 02 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Decide which agent tool calls need human approval</title><link>https://airoweb.com/blog/agent-tool-approval-matrix/</link><guid isPermaLink="true">https://airoweb.com/blog/agent-tool-approval-matrix/</guid><description>A practical approval matrix for AI agents, MCP tools, and connected workflows that can read data, write records, or trigger external actions.</description><pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Review AI pilots before they become business process</title><link>https://airoweb.com/blog/ai-pilot-to-process-review/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-pilot-to-process-review/</guid><description>A practical review for deciding which AI pilots should scale, stay limited, or stop.</description><pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Fable 5 is back. Do not build as if access is permanent.</title><link>https://airoweb.com/blog/claude-fable-5-limited-availability/</link><guid isPermaLink="true">https://airoweb.com/blog/claude-fable-5-limited-availability/</guid><description>Claude Fable 5&apos;s launch, suspension, and capped return are a practical warning about frontier-model dependencies.</description><pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate></item><item><title>A better prompt for building an Obsidian vault as agent memory</title><link>https://airoweb.com/blog/obsidian-vault-prompt-for-agent-memory/</link><guid isPermaLink="true">https://airoweb.com/blog/obsidian-vault-prompt-for-agent-memory/</guid><description>A practical teardown and rewrite of a prompt for turning local notes, source traces, and connector context into a validated Obsidian vault for LLM workflows.</description><pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Before buying AI tools, write the operating model</title><link>https://airoweb.com/blog/ai-adoption-operating-model/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-adoption-operating-model/</guid><description>A practical operating model for deciding where AI belongs, who reviews the work, and how teams avoid tool sprawl.</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate></item><item><title>A quick review checklist for recurring AI workflows</title><link>https://airoweb.com/blog/ai-workflow-review-checklist/</link><guid isPermaLink="true">https://airoweb.com/blog/ai-workflow-review-checklist/</guid><description>A practical review checklist for deciding whether an AI-assisted workflow is ready for everyday operations.</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate></item><item><title>MCP catalogs can wait. The workflow cannot.</title><link>https://airoweb.com/blog/using-mcps-internal-workflows/</link><guid isPermaLink="true">https://airoweb.com/blog/using-mcps-internal-workflows/</guid><description>Why airoweb should explain concrete MCP workflows, permission boundaries, and review gates before publishing a catalog of integrations.</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate></item></channel></rss>