Historical source material recovered on 2026-09-09. Capture dates and source limitations apply; recovery is not fresh verification.

AI Coding Agents

AI coding agents are systems that can plan, edit code, run tools, execute tests, and participate in software delivery beyond simple code completion.

Key facts

• BMAD positions itself as a workflow and methodology layer for coding assistants that can follow structured roles, plans, and QA loops ^[src-184].

Related

• Agentic Coding
• Agentic Software Development Lifecycle
• Codex CLI

2026-07-07 CLI landscape update

• The mid-2026 CLI coding-agent survey counts 35 actively maintained CLI agents and treats Claude Code, Codex CLI, Gemini/Antigravity CLI, Grok Build, Mistral Vibe, Kimi Code, Qwen Code, Aider, OpenCode, Cline, Goose, and related tools as one increasingly standardised category ^[src-208].
• For AI Watch, the important transition is from IDE plugin UX to shell-native automation that can run in CI, SSH sessions, and machines without a GUI ^[src-208].

Source references

• ^[src-184] bmad-code-org – "BMad Method: Breakthrough Method for Agile AI-Driven Development" (2026-06-22)
• ^[src-208] arcbjorn – "State of CLI Coding Agents, Mid-2026" (2026-07-04)

2026-07-17 Grok Build update

• Grok Build provides a concrete open-source harness architecture: separate crates for TUI, runtime, tools, workspace/VCS execution, checkpoints, configuration, MCP, skills, plugins, hooks, sandboxing, and headless operation ^[src-244].
• Agent Client Protocol support makes the same runtime embeddable in editors, while headless mode supports scripting and CI ^[src-244].

Historical recovered snapshot — 2026-08-21

Preserved from ai-watch-src380-392-enterprise-agents-20260821. Documentary recovery on 2026-09-09; claims below retain their historical source scope and were not freshly verified. Current synthesis above remains authoritative.

• BMAD positions itself as a workflow and methodology layer for coding assistants that can follow structured roles, plans, and QA loops ^[src-253].

• ^[src-253] bmad-code-org – "BMad Method: Breakthrough Method for Agile AI-Driven Development" (2026-06-22)
• ^[src-208] arcbjorn – "State of CLI Coding Agents, Mid-2026" (2026-07-04)

• Grok Build provides a concrete open-source harness architecture: separate crates for TUI, runtime, tools, workspace/VCS execution, checkpoints, configuration, MCP, skills, plugins, hooks, sandboxing, and headless operation ^[src-244].
• Agent Client Protocol support makes the same runtime embeddable in editors, while headless mode supports scripting and CI ^[src-244].

Enterprise distribution and shared review update (2026-08-21)

• Antigravity's inclusion in Gemini Enterprise moves coding-agent competition into licence distribution, shared token capacity, project budgets, identity, policy and audit across desktop, CLI and IDE surfaces ^{[src-382] [src-383]}.
• Slack Code adds a collaborative work surface where teams can inspect an agent conversation, diffs and live previews, give feedback and approve work before shipping ^[src-386]. Shared visibility improves reviewability but does not replace repository history, CI or deployment controls.

• ^[src-382] Google Cloud — Antigravity enterprise expansion announcement (2026-08-21)
• ^[src-383] Google — Antigravity enterprise documentation (inspected 2026-08-21)
• ^[src-386] Jess Weatherbed / The Verge — “Slack is launching collaborative vibe-coding channels” (2026-08-20)

• Google Cloud's modernisation tutorial makes repository inventory, executable tests, reviewed plans, small diffs and final acceptance checks the control plane around a coding agent ^[src-385]. The source is a demonstration, not measured production evidence.

• ^[src-385] Google Cloud Tech — legacy-code modernisation tutorial (2026-08-20)