Field research and feature analysis for AI coding agents — sandboxing, orchestration, plugins, community tooling, SDLC patterns.
Live: knowledge graph · site
A continuously-updated research catalog for making adopt / defer / skip decisions about AI coding agents and their ecosystems before building production systems on them. Every analysis follows one structure: What it is → How it works → Adoption decision → Action items.
- Claude Code (Anthropic-native) — agents/skills, CI/sandboxing, context/memory, configuration, plugins, MCP connectors
- Non-CC agents & infrastructure — JetBrains Air, DeerFlow, Goose, Codex, Devin, frameworks, orchestrators
- Community ecosystem — skills, plugins, tooling, domain-specific CLAUDE.md patterns
- SDLC / lifecycle management — agentic SDLC patterns, OSS ALM landscape
- Cross-repo learnings and an auto-generated cumulative index of agentic-AI papers
- Self-currency — four cron monitors open triage PRs when upstream sources change
Full directory map and the monitor pipeline live in docs/architecture.md; the research goals and user stories are in docs/UserStory.md; browse the analyses under docs/.
Doc tooling installs user-locally (zero sudo, into ~/.local/bin) and runs through the Makefile:
make setup_all # lychee + Node.js + markdownlint-cli2 + actionlint + shellcheck
make lint # links (lychee) + markdown (markdownlint-cli2) + actions (actionlint)
make test # stdlib unit tests for scripts/ + .github/scripts/lib/ modules
make help # all recipes, grouped by sectionSee CONTRIBUTING.md for document standards and CHANGELOG.md for history.
Make informed decisions about coding agents before committing production systems to them. The catalog tracks Claude Code, JetBrains Air, DeerFlow, Goose, Codex, Devin and others — plus the surrounding plugins, observability, SDLC patterns, and cross-repo learnings distilled from live development across the qte77 ecosystem.
- An Open Agentic Coding Harness — the write-up this catalog feeds
- cc-recursive-team-mode — recursive CC subprocess spawning, session-artifact research
- coding-agent-eval — hands-off coding-agent comparison harness
- multi-tasking-quality-benchmark — WakaTime activity vs code-quality correlation
- Agents-eval — origin: these analyses began here, to inform a multi-agent evaluation framework