BC Observatoryunofficial, made by waldo
videoAL Developmentcommunity - not Microsoftreviewed - checked by Opus

Episode 513: No Windows, No IDE, No Problem: BC Development from the Terminal

Dynamics Corner episode 513 covers terminal-based, AI-driven Business Central AL development with Claude Code on Linux, plus a third-party AL code review service built on an ISO 5055-based analyzer. It also shows a prototype of Business Central running in Linux Docker containers.

Open in Claudew2P3WskFaCA.md
  1. 0:00 Introduction and keeping up with rapid change
  2. 1:55 Challenges of staying current with new features
  3. 4:06 Imposter syndrome and rapid feature introduction
  4. 6:20 Introducing Stefan Moran and code review service
  5. 7:48 AI-assisted development and code review needs
  6. 9:51 Business Central unique constraints and framework considerations
  7. 11:29 Code review service methodology and analyzer framework
  8. 14:49 Two-level code review offering and expert service
  9. 15:51 Lint cop roots and risk of vibe coding in business applications
  10. 17:27 Risk assessment for AI-assisted development in production
  11. 19:15 Cloud code workflow and AL instruction context management
  12. 21:15 Plugin architecture and Flow Next agent workflow approach
  13. 23:21 File-based context strategy for multi-agent development
  14. 25:16 AL instructions development and iterative improvement process
  15. 27:49 Evolution from commands to skills in agent orchestration
  16. 30:20 Re-engineering frameworks and staying current with features
  17. 31:49 Finding use cases: problem-driven learning approach
  18. 34:07 Building momentum through small tasks and workflows
  19. 35:38 AI-generated code workflow and minimal VS Code usage
  20. 37:45 Review process and pragmatic tool selection
  21. 39:44 Parallel processing and terminal-based development workflow
  22. 42:25 Cloud code integration and Linux development
  23. 43:59 Cross-platform development: Mac, Linux, and CLI
  24. 45:29 Handling reports and Windows VMs in Linux environments
  25. 47:25 Custom Bash scripts and publishing workflow
  26. 49:38 Docker containers and Linux Business Central project
  27. 51:52 Testing and standalone code execution
  28. 53:54 Performance and pipeline benefits
  29. 55:35 AI-assisted development and testing
  30. 58:25 Test-driven development with Linux
  31. 1:00:43 Context switching and mental load
  32. 1:03:46 Future of code generation and language independence
  33. 1:07:05 Human-readable code and AI optimization
  34. 1:09:05 Developers as early adopters of AI tools
  35. 1:10:50 AI thinking, interaction design, and hallucination
  36. 1:13:00 Statistics, LLMs, and the history of computing
  37. 1:15:00 Linux Business Central container project and GitHub
  38. 1:16:45 Closing remarks and presenter contact information
video Dynamics Corner podcast · 2026-04-14Watch on YouTube

Overview

The episode has two guests’ topics. First, Stefan Moran describes a code review service for AL code. It has a free automated assessment from a proprietary analyzer and a paid detailed review. The analyzer maps about 170 ISO 5055 rules to AL, adds Business Central specific rules, and runs roughly 1500 automated tests. The motivation is that more AI-generated solutions are going into production and some people may want an outside opinion on quality.

Second, the presenter describes a workflow where nearly 100 percent of code is AI-generated, using Cloud Code from the terminal with little VS Code. It uses AL instructions in a plugin, file-backed sub-agents that return only summaries, bash wrappers for compiling and publishing, and test-first development with an approval gate. He also demonstrates a fragile prototype that runs Business Central and its tests in Linux Docker containers, aimed at faster pipelines. The talk ends with speculation on language-independent code generation.

Key points

  • The code review service has a free automated assessment and a paid detailed review. The analyzer maps about 170 ISO 5055 rules to AL, leaves out rules such as memory exploits, and adds custom Business Central rules.
  • The speaker says he is approaching 100% AI-generated code but reviews and tests everything. He finds cleanup with existing quick fixes and community tools faster and more reliable than having AI do it.
  • Context management approach: sub-agents write results to files and return only a summary to the main agent, so sessions can be stopped and resumed from disk.
  • AL instructions are packaged as a plugin and are refined by trial and error on real projects. Commands are described as a little deprecated in favor of skills, which can force a sub-agent to be spawned.
  • Helper scripts: an al-compile bash wrapper around the AL compiler, and a bc-publish command that uses curl against the deployment API.
  • Test-driven approach: tests are written and must fail before code is generated, and a manual approval gate follows. The speaker calls this the only way he found to trust AI output without reviewing every line.
  • The Linux Business Central container project runs a service tier with SQL Server in Docker, with a Linux-compatible HTTP replacement. The speaker calls it fragile and still in progress, and it is public on his GitHub. Windows users should prefer BC container helper.

Features

Feature Status At
Code review service status not stated 7:30
AL code analyzer status not stated 12:23
AI-assisted development in Business Central status not stated, demoed 7:30
Cloud Code AL development workflow status not stated, demoed 19:15
Flow Next file-backed agent workflow status not stated, demoed 21:15
Multi-agent context management strategy status not stated, demoed 24:05
Iterative AL instruction refinement status not stated 25:16
Commands framework status not stated 28:49
Skills framework status not stated 29:07
Problem-driven feature adoption status not stated 32:23
Parallel project development status not stated 40:29
Terminal UI and File Explorer Navigation status not stated, demoed 41:35
Arch Linux Development Environment status not stated, demoed 43:33
Windows VM for Reporting status not stated, demoed 45:29
Windows VM Running in Linux Docker Container status not stated, demoed 45:40
AL Compiler Bash Script Wrapper status not stated, demoed 47:44
Publishing via Curl CLI Command status not stated 48:41
Docker Containers for Business Central on Linux status not stated, demoed 49:57
Standalone Codeunit Execution Without Service Tier status not stated, demoed 51:32
C# HTTP Implementation Replacement for Linux Compatibility status not stated, demoed 52:20
Local Linux Service Tier Execution status not stated, demoed 53:15
Automated CLI-based Test Execution status not stated, demoed 56:30
Test-driven Development with AI Code Generation status not stated 55:49
OData-based Test Exposure status not stated 1:00:43
AI-driven code generation and language abstraction status not stated 1:05:02
Machine-optimized programming languages status not stated 1:06:53
Linux Business Central container project status not stated 1:14:11
Business Central source code history repository status not stated 1:15:38

AL objects mentioned

As heard in the captions. A name that matches one object page by exact type and name links to it; the others stay as named.

  • other “ISO 5055 standard” at 13:06
  • other “analyzer” at 12:38
  • other “lint cop work” at 15:37
  • other “air cops initiative” at 15:37
  • codeunit “code unit” at 51:52
  • other “cloud code” at 52:20
  • other “BC container helper PowerShell scripts” at 59:30
  • other “management API” at 1:00:43
  • other “AL objects” at 1:13:08

Not found in BC28-30: codeunit “code unit”.

Quotes

  • 3:33 “I I want to keep up and I feel I I have to keep up with everything”
  • 3:44 “if you would skip a few things and a few obsolet cycles of new features, you would probably still need to get up to”
  • 8:15 “more and more solutions are getting used and deployed and everything put into production which are really nearly 100”
  • 8:33 “maybe there would be a need for people to have an opinion or to to buy an opinion basically from somebody who is not”
  • 10:40 “we are a little bit more constrained with things that are uh outside of the coding frame a little bit with the runtime with”

Disclaimers in the video

  • 13:43 subject-to-change: working on that and the review basically is a a a sum of all of that basically every everything gets …
  • 25:00 subject-to-change: Memory feature existence and function discussed as unclear and not fully understood
  • 28:37 subject-to-change: I’m I’m in the middle of reworking it again. Also for the stream approach with the containers and everything
  • 50:44 preview: it’s almost working, but pretty fragile if I may say so
  • 53:25 subject-to-change: Now I’m still working on making this a little bit more stable
  • 53:25 preview: I’m still working on making this a little bit more stable, but I have like automated tests where like 270 …
  • 1:14:28 preview: It’s already public if you want to have a look on my on my GitHub profile. should be the top …

Presenters (as heard): Stefan Moran, Chris, Brad, Claus Khan.