flow

mcp
Security Audit
Warn
Health Warn
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Pass
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Aixle Flow — orchestrate coding agents through durable, inspectable workflows.

README.md

Aixle Flow

The team-level control plane for AI coding agents.
Put a card on the board, an agent picks it up in an isolated container,
your team sees the run, the cost, and the output — without anyone
opening a terminal.

License: Apache 2.0
Built with Rails
React 19
Status: Pre-1.0

CI
Backend coverage
Frontend coverage

Aixle Flow — a card is dragged into an automated column; the bound workflow runs Plan, Implement, Run tests and Open PR across Claude Code, Codex and Cursor CLI, each in its own container, and the sessions table shows the tokens and cost each step spent

Why Aixle Flow

Cursor and Claude Code are personal superpowers. They run on your
laptop, in your terminal, for your task. Once more than one
person on a team starts delegating work to agents, the wheels come off:

  • Nobody knows what's running, where, or whether it's stuck.
  • There's no shared trail of what an agent did or how much it cost.
  • Setup is per-developer, so the team uses agents five different ways.

Aixle Flow is the layer above the agent — the Kanban board, the workflow
graph, the audit log, the cost meter. The agent stays great at coding;
Aixle Flow makes it usable by a team.

It is the only open platform we know of that combines:

  • A board with column → workflow bindings (drop a card, trigger a run)
  • A DAG workflow engine with retries, approval gates, and parallel steps
  • Container-isolated agent execution (Claude Code, Cursor CLI, Codex, Gemini CLI)
  • MCP-native tool integration
  • Per-session cost tracking in tokens and cents
  • Temporal under the hood for durable, long-running runs

Quickstart

Three commands to a working app on http://localhost:4000.

git clone https://github.com/AixleHQ/flow.git && cd flow
cp .env.example .env.development   # fill in OAuth secrets if you want SSO
make setup                          # builds containers, installs deps, seeds DB

Then in two terminals:

make up       # terminal 1 — web, db, redis, traefik, temporal
make worker   # terminal 2 — Temporal worker

Open http://localhost:4000 and sign in.

5-minute setup is on the roadmap. Right now make setup builds
agent images which takes longer the first time. See
ROADMAP.md for the planned docker compose up
single-command flow.

Need more detail? See docs/quickstart.md.

Documentation

Three tiers, each independently navigable:

Tier For Where
Quickstart First-time users, < 5 min docs/quickstart.md
User Guide Common workflows & gotchas docs/user-guide/
Reference Full API / CLI / config surface docs/reference/

Roadmap

What's planned and what's in flight — see ROADMAP.md.
Issues tagged good first issue are a good place to start
contributing.

Tech Stack

  • Backend — Ruby on Rails 8, PostgreSQL, Redis, Temporal
  • Frontend — React 19, Inertia.js, Mantine 9, Vite, TypeScript
  • Agents — Claude Code, Cursor CLI, OpenAI Codex CLI, Gemini CLI
    (each runs in its own Docker container)
  • Protocols — MCP for tools, ActionCable for live updates
  • Dev environment — Docker Compose, one Makefile

Inspired by

Aixle Flow stands on the shoulders of the open agent-and-workflow
ecosystem. A few projects shaped how we think:

  • n8n — workflow automation can be open,
    code-friendly, and priced per run rather than per step.
  • Dify and Langflow
    — visual, open-source orchestration of LLM agents and pipelines.
  • Paperclip — treating a roster of agents like an org, with budgets,
    governance, and per-agent cost as first-class concerns.
  • Cursor — background agents that run in
    isolation and are triggered by events, not just a human at a keyboard.
  • Temporal — durable execution as the right
    foundation for long-running, retryable agent runs.
  • Model Context Protocol — an open
    standard for giving agents tools.
  • The Kanban board — work
    lives in columns anyone on the team can move.

Where each inspired a category, Aixle Flow's own contribution is to make it
board-native and team-first: drop a card, an agent runs in a container,
and the whole team sees the run, the trail, and the cost.

Contributing

We welcome PRs, issues, and discussion.

  1. Read CONTRIBUTING.md.
  2. Pick up a good first issue or open a discussion before
    starting on something larger.
  3. Run make check before opening a PR.

License

Aixle Flow is released under the Apache License 2.0 — a permissive
license that lets you use, modify, self-host, and commercially deploy it freely,
including in closed-source products, as long as you preserve the license and
attribution notices.

See LICENSE for the full terms, NOTICE for attribution,
and THIRD-PARTY-LICENSES.md for the licenses of
bundled dependencies. "Aixle" and "Aixle Flow" are trademarks — see
TRADEMARK.md.

Contributions are accepted under the Apache 2.0 license and require both a
Developer Certificate of Origin (DCO) sign-off and a signed Contributor License
Agreement (CLA) — see CONTRIBUTING.md.

Reviews (0)

No results found