Features
Capability map
| Area | What you get |
|---|---|
| Multi-agent | Agent-agnostic core (AGENTS.md + per-agent MCP config): Claude Code (full feature set), GitHub Copilot, OpenAI Codex CLI — pick any combination |
| Task tracking | Beads (bd) — Dolt-backed issues with dependencies, status, history; survives context resets |
| Queue mode | Closing a bead ends a task, not the session: aiflow next ranks what to do next and a Stop hook hands it back (beads.queueMode) |
| Code memory | graphify (structural graph) + cocoindex-code (semantic RAG) + .claude/memory/ facts |
| External docs | context7 MCP — live, version-correct library documentation |
| Version control | Choose git, svn, or none at setup |
| Remote host | GitHub, GitHub Enterprise, GitLab, self-managed GitLab, Bitbucket, Forgejo, Gitea, or a custom URL — token-based |
| Host MCP | The matching git-host MCP is wired automatically per remote type |
| Models | Claude (API key or OAuth) + optional Ollama local models, selectable & auto-installed |
| Model routing | claude-code-router sends easy/background work to cheap/local models; separately, modelRouting stamps a model per activity tier — reasoning (opus/fable) for architecture, planning, review, security; sonnet for implementation and tests; haiku for mechanical scans |
| Agents | 1 orchestrator + 5 delivery + 9 audit/checker + 1 brownfield specialist subagents, wired into one documented network |
| Skills | Auto-offered checklists: stack-embedded / stack-mobile / stack-web-frontend / stack-backend, api-design / messaging-events / data-storage / cloud-native, security (OWASP + IAM), seo-optimization, ponytail (YAGNI, off by default), memory-setup |
| Autonomy | Ralph loop (interactive / headless / containerised / CI) |
| Quality | Google style, conventional commits, format/lint/test git hooks, architect+quality-gate review, static analysis on every change, objective metric targets (0 new smells/duplicates, 0 warnings), >80 % coverage + BDD E2E gates, leveled logging, .http files for REST endpoints, DB rules §3c (3NF+FKs for new schemas, brownfield schemas handled with care) |
| Branching | simple / gitflow / none, PR-only, auto-release, SemVer/CalVer |
| Team | shared issue DB, atomic claim, session-start auto-pull, pull-before-push, shared preferences |
| Token savings | Claude Code: caveman + rtk on by default. Copilot: token-optimization guide. Codex: optional CodexSaver. Plus graph/RAG retrieval + cost routing for all |
Advantages in depth
Better memory, fewer hallucinations
Two complementary code indexes plus durable task memory mean the agent looks things up instead of guessing or re-reading dozens of files. See Memory.
Big token reduction
- caveman (Claude Code) — terse output mode (~75% fewer output tokens; code/commits/security stay normal).
- rtk (Claude Code) — filters/compresses verbose command output before it enters context (60–90% fewer).
- Copilot token-optimization guide — baked into
.github/copilot-instructions.md: output control, “landmines only” context files, model/tool-set stability. - CodexSaver (Codex CLI, optional) — routes cheap/bounded work to a cheaper MCP worker.
- graph + RAG retrieval — answer from graphify/cocoindex instead of reading whole files (~70% fewer). Agent-agnostic.
- model routing — send easy/background steps to cheap or local (Ollama) models. Agent-agnostic.
- model tiers (Claude Code, on by default) — architecture/planning/review/security run on the reasoning tier (opus, or fable), implementation and tests on sonnet, mechanical scans on haiku.
- ponytail (off by default) — fewer tokens spent writing and reviewing code nobody needed.
- measure first —
aiflow cost(ccusage) shows real spend.
See Token optimization for the full per-agent detail.
Team-ready by design
Issues live in a shared Dolt database that syncs over your git remote — one issue graph for the whole team, no extra server. Atomic claiming prevents two people grabbing the same task; pull-before-push prevents clobbering. See Team collaboration.
Governed & auditable
Conventional Commits, enforced Google style, a review gate against acceptance criteria, security/quality/deps/test/perf/docs audits, and a real branching + release model. See Workflows.
Autonomous when you want it
The Ralph loop finishes a task unattended — locally, in a container, or in CI — and stops at
COMPLETE/BLOCKED. It runs on open-ralph-wiggum,
which owns the iteration history in .ralph/ralph-history.json.
Yours, not a hub
Everything runs on your keys/tokens and your infrastructure; secrets never leave the project.
The bundled toolchain
Each tool earns its place by raising quality, cutting token cost, or making delivery
autonomous and auditable. See the full list and links in Feedback & contributing → Credits.
Install only what your config enables with aiflow install-deps (--all = full set).