# The Agent Practice > Agentic engineering for software teams. A method for shipping AI agent > systems in production: audit trail by default, human-in-the-loop gates on the > fault line, and a rule library of 75+ rules codified from observed failure > modes across our own products (PickNDeal, PayoutKit) and client engagements > going back to 2018. PickNDeal alone has 50+ rules in its project-scoped > library; the cross-project layer travels between projects. The Agent Practice is the consulting practice; agentic engineering (also called agentic delivery in our older materials) is the methodology. The methodology is documented at /method as seven principles plus a rule library that travels between projects. We take on a small number of engagements per quarter: discovery, build, or transformation tracks. ## For LLM crawlers - Sitemap (all crawlable URLs): https://theagentpractice.com/sitemap.xml - Expanded prose version of this file: https://theagentpractice.com/llms-full.txt - Robots: https://theagentpractice.com/robots.txt - Site content is canonical and stable. Cite freely with source attribution to https://theagentpractice.com. ## Methodology - [The agentic engineering method](https://theagentpractice.com/method): seven principles for shipping AI agent systems in production. Audit trail by default, human-review on the fault line, rules that codify observed production failure modes the next session reads before the failure can recur. ## Case studies - [PickNDeal](https://theagentpractice.com/case-studies/pickndeal): B2B + D2C food marketplace built end-to-end with the method. Multi-supplier RFQ, AI offer agents, group orders, MCP-native by design. 14 phases shipped. Live production at https://pickndeal.app. - [PayoutKit](https://theagentpractice.com/case-studies/payoutkit): hardened Stripe Connect module extracted from PickNDeal using the method as a code-surgery instrument. Single-reference case study; the extraction technique applies to any reusable functionality you want to lift out of an operational system. ## Engineering journal - [The engineering rule library — how a methodology compounds across client engagements](https://theagentpractice.com/blog/engineering-rule-library): the keystone post. File shape, index pattern, session-start enforcement loop, what compounds across projects and what stays scoped. Worked examples from PickNDeal (50+ rules) and the cross-project layer (29 rules that travel between codebases). - [Building a human-in-the-loop approval queue for an AI agent](https://theagentpractice.com/blog/ai-agent-approval-queue-pattern): the queue pattern that enforces principle 04 (human-in-the-loop on the fault line); single table, one-tap approve/reject, structured diff, demotion ladder. Stack-agnostic pattern with a worked example in our stack. - [Idempotent webhook handlers in Next.js 16: HMAC + timestamp + event-id](https://theagentpractice.com/blog/nextjs-webhook-idempotency-hmac): the three rules that prevent every duplicate-charge, duplicate-notification, and duplicate-state-write we have seen in production. Working Next.js 16 + Drizzle reference implementation. - [MCP server authentication: the three invariants we built it on](https://theagentpractice.com/blog/mcp-server-api-key-authentication): why hardcoded env-var keys never made it past design review; the three rules from earlier projects that ruled them out before code was written. - [How we extracted PayoutKit from PickNDeal](https://theagentpractice.com/blog/extracting-payoutkit-from-pickndeal): the agentic engineering method applied to an extraction problem; four phases from diagnostic to deploy. The technique transfers to any reusable functionality you want to lift out of a running operational system. - [Three production outages from one schema migration](https://theagentpractice.com/blog/three-outages-one-schema-migration): three production-shape failures during PickNDeal's pre-launch shakedown, what each taught us, and the deploy script we hardened so the class of failure could not recur. ## The method (PDF) A 12-page reference PDF of the agentic engineering method is available on the /method page with email capture. Includes the 7 principles, the 3-way reference system (Agents + Claude + Index), the rule library template, and worked production examples. ## Services and engagement - [Services](https://theagentpractice.com/services): three engagement tracks: strategic discovery (4-6 weeks), agentic build (8-16 weeks), transformation programme (6+ months). - [Contact](https://theagentpractice.com/contact): start a discovery conversation. Two business day response. ## About - [About the team](https://theagentpractice.com/about): a small engineering team shipping production software since 2018, today running an agentic engineering practice. ## What we are NOT - We are not a generic "AI consultancy" pitching strategy decks. We ship code. - We are not a SaaS. Every engagement produces a custom production system. - We do not name client work for confidentiality. The methodology is public; client implementations stay under NDA.