Capability 01 of 04 · Development
Coding Agent Usage
How your team uses AI coding assistants - from autocomplete to autonomous agent fleets.
5
maturity levels
15
practices
15
guides
L1 · Stage 01Assisted
Practices - what it looks like3 guides
- Copilot autocompleteHow to use IDE autocomplete as your first step into AI-assisted development.guide→
- Chat in sidebar, ad-hoc questionsHow to use the AI chat panel for one-off code questions and explanations before any systematic workflow exists.guide→
- Agent runs without codebase contextUnderstanding why AI tools at L1 see only what you show them - and why that's the core limitation the entire maturity journey addresses.guide→
Criteria - what to measure
- 01At least one AI coding assistant (Copilot, Cursor, Claude Code) is installed and active for at least one developer
- 02AI autocomplete or chat is used at least once per week by the team
- 03Developers have access to AI chat in their IDE sidebar
- 04Team has experimented with AI-assisted code generation on non-critical tasks
L2 · Stage 02Delegated
Practices - what it looks like3 guides
- Agent in IDE; autonomy set by a written rule, not a per-prompt click (Claude Code auto mode became the default Aug 14 behind a classifier - `permissions.deny` and `ask` still override it)How to run an IDE agent whose autonomy is set by a written, version-controlled permission ruleset rather than a per-prompt click, so multi-file work runs without constant confirmation interrupts.guide→
- An agent instruction file ships with every active repositoryAn agent instruction file ships with every active repository, teaching AI tools that project's conventions, patterns and constraints - the single highest-leverage action at L2.guide→
- Copilot + Claude Code in parallelHow to use inline autocomplete and an agentic CLI tool simultaneously, each at the granularity it handles best.guide→
Criteria - what to measure
- 01Agents operate in multi-step agentic mode (edits without per-step approval)
- 02At least one agentic IDE (Cursor, Windsurf, or Claude Code) is used by 50%+ of the team
- 03CLAUDE.md, .cursorrules, or equivalent agent instruction file exists in 100% of active repositories
- 04Developers use two or more AI tools in parallel (e.g., Copilot + Claude Code)
- 05Agent instruction files are reviewed and updated at least quarterly
- 06Agent autonomy is bounded by a written permission ruleset (deny/ask) committed to the repo, not by per-prompt clicking
L3 · Stage 03Systematic
Practices - what it looks like3 guides
- Code is written to be read by agents: explicit over implicit, obvious over cleverCode is written to be read by agents: explicit over implicit, obvious over clever, with the conventions that govern it stated precisely rather than absorbed by osmosis.guide→
- Rules files per-team/per-repoHow to evolve from a single project-level CLAUDE.md to a layered system of context files tailored to each team's tech stack and conventions.guide→
- CLI agents as primary (Claude Code with Opus 5, Codex on GPT-5.6, Grok 4.6, Meta Muse Code) with open-weight workers (GLM-5.3, Qwen3.8-Max, DeepSeek V4 Pro) for executionHow shifting from IDE plugins to CLI-based agents makes AI a programmable, scriptable part of your development workflow rather than a typing assistant.guide→
Criteria - what to measure
- 01Coding conventions are written as explicit, agent-parseable rules (not implicit tribal knowledge)
- 02Per-team or per-repo rules files exist and are maintained with code review
- 03CLI agents (Claude Code, Codex) are the primary coding interface for 50%+ of feature work
- 04Agent usage is tracked per developer and per repository
- 05Agent instruction files follow a standardized template across the organization
L4 · Stage 04GovernedMost teams aim here
Practices - what it looks like3 guides
- Scheduled / unattended agents + model routing: frontier plans, cheap executes - re-costed on a schedule, because the cheap tier inverted in four weeks (Cursor Router, Amp; Stripe Minions)How to launch AI agents that run to completion autonomously in a sandbox - writing code, running tests, fixing errors and opening a PR without supervision - under a run-status taxonomy and a model-routing policy you re-cost on a schedule.guide→
- Slack/CLI/Web/PagerDuty invocation → PRHow to trigger AI agent tasks from natural language interfaces - a Slack message, a CLI command, a web form - and have the agent autonomously produce a pull request.guide→
- 3-5 parallel agents per developer + merge queues for agent fleets; the ceiling is orchestrator context pollution, not token cost (cap batches at 2-4, no concurrent repo-wide git ops)How to shift from sequential AI assistance to managing multiple concurrent agent instances - transforming the developer's role from implementer to orchestrator, with the ceiling set by orchestrator context pollution rather than by token cost.guide→
Criteria - what to measure
- 01Unattended agents execute tasks without developer presence
- 02Agents are invocable from at least two channels (Slack, CLI, Web, PagerDuty)
- 03Developers routinely run several agent sessions concurrently, against a documented span-of-control limit
- 04Agent task completion rate without human intervention exceeds 60%
- 05Agent invocation produces a PR within a defined SLA (e.g., under 30 minutes for standard tasks)
L5 · Stage 05Self-improving
Practices - what it looks like3 guides
- Multi-agent orchestration (Claude Code dynamic workflows - the Bun-in-Rust model, Gas Town / custom)How to build systems where specialized agents collaborate - a planner decomposes tasks, workers execute them, and reviewers validate results - to handle complex engineering tasks end-to-end.guide→
- Planner → Worker hierarchyHow to structure a two-tier agent architecture where a planner decomposes engineering tasks and workers execute them in parallel - the canonical L5 pattern for complex autonomous development.guide→
- Fleet size bounded by compute and review capacity, not by toolingThe frontier of AI-assisted development: massive agent parallelization where hundreds of concurrent agents produce thousands of commits per hour on a single codebase.guide→
Criteria - what to measure
- 01Multi-agent orchestration system (planner-worker hierarchy) is in production
- 02The agent fleet scales past what a single team could supervise, bounded by compute and review capacity rather than by tooling
- 03Agent fleet produces 1,000+ commits per week without manual dispatch
- 04Planner agents decompose epics into tasks and assign to worker agents autonomously
- 05Agent fleet self-recovers from failures without human escalation for 90%+ of error cases
From the Field
Recent releases, projects and discussions the AI Radar classified into this capability.
release
anthropics/claude-code
article
Under-utilized features in Claude Code
discovered
claude code source map v2.1.88
article
Free open source AI Editor
discovered
MrsEWE44/musicDownload
release
kodustech/kodus-ai
discovered
A Chrome extension for orchestrating local AI-agent team workflows across web AI services.
article
Ask HN: What coding agents are you using?
Which level is your team at in Coding Agent Usage?
The criteria above are what we check in a facilitated assessment. Run it yourself to place this capability, see which gates you have passed, and compare it against the other 3 in Development.
Start the assessment →