Skip to content
Sign in
  • Browse all toolsFilter by category and pricing
  • Browse categoriesWriting, code, image, audio and more
  • Compare toolsPut up to three side by side
  • Read the blogGuides and comparisons

Ruflo

Open-source agent orchestration for Claude Code, Codex and MCP clients, with swarms, memory and workflow controls.

Ruflo is an open-source agent harness that adds coordinated swarms, memory, task routing and execution controls to Claude Code and Codex. Ruflo serves developers building agent-assisted coding workflows and is available under the MIT license through a public repository, with CLI, plugin and MCP installation paths.

Updated 9 Nov 20266 sections

Ruflo website screenshot

What is Ruflo?

Ruflo is an open-source agent harness for developers working with Claude Code, Codex and other MCP clients. Ruflo supplies the execution layer around a model, adding tools, memory, task coordination and controls. The repository identifies Reuven Cohen as its creator and says the project was originally called Claude Flow.

Ruflo separates model output from the surrounding workflow machinery. The model writes, while the harness organizes tasks, carries context between steps and helps check results. Ruflo's documented architecture connects its CLI or MCP interface to routing, swarms, agents, memory and configured model providers.

What Ruflo can do

Ruflo combines agent coordination with memory and workflow inspection. Ruflo's hooks can route tasks, learn from successful patterns and coordinate agents after project initialization. The console exposes execution details, while optional plugins add reusable workflows, scheduled work and communication between agents on different hosts.

  • Ruflo coordinates multiple agents as a swarm.
  • Ruflo retains memory across sessions.
  • Ruflo supports vector retrieval and RAG integration.
  • Ruflo enables agent federation across machines.
  • Ruflo displays agent logs, results, tokens and costs.
  • Ruflo supports reusable task templates and scheduled background tasks.

How Ruflo works

Ruflo's project setup starts with npx ruflo@latest init in a project folder. Ruflo adds agents, hooks, memory and an MCP server; Claude Code picks up that configuration on its next start. Users can then give Claude Code a concrete task, such as reviewing a repository and writing tests.

Ruflo also supports a plugin-only path inside Claude Code without adding project files. Ruflo's core plugin registers its own MCP server, but other plugins generally do not. Codex and other MCP clients can connect through npx -y ruflo@latest mcp start, and the console can run independently with npx ruflo tui.

Who Ruflo is for

Ruflo is for developers who want coordinated agents around coding tasks rather than an isolated model conversation. Ruflo's documented examples cover repository review, test writing and mission creation. The console also supports users who need to inspect failures, compare runs and retain approval authority over agent actions.

  • Ruflo can coordinate repository review and test-writing tasks.
  • Ruflo can support reusable multi-step development workflows.
  • Ruflo can help users inspect logs and triage failed runs.
  • Ruflo can connect collaborating agents on different machines.

Ruflo pricing and access

Ruflo is distributed as an open-source project under the MIT license through its public repository. Ruflo's documented access paths include project initialization, Claude Code plugins, direct MCP configuration and a standalone terminal console. The supplied page does not list subscription plans or prices for model-provider usage.

Ruflo strengths and limitations

Ruflo brings coordination, persistent context and execution inspection into the same developer workflow. Ruflo also exposes permission levels and approval settings for console actions, with a take-back-control option. Network, spending and destructive actions require confirmation even when automatic approval is enabled, according to the documented console controls.

  • Ruflo runtime mods are not sandboxed and run with account permissions.
  • Ruflo's plugin path does not install the full CLI path's hooks.
  • Ruflo's local retrieval and learning require embeddings and configured learning modules.
  • Ruflo mission creation does not mean the requested feature has been implemented.

Ruflo FAQ

What is Ruflo?

Ruflo is an open-source agent harness for Claude Code and Codex that adds coordinated swarms, memory, task routing and execution controls. Ruflo also supports other MCP clients and provides a console for inspecting workflows, agent logs and results.

Is Ruflo free and open source?

Ruflo is available as an open-source project under the MIT license in a public repository. Ruflo's supplied documentation describes CLI, plugin and MCP access paths but does not list subscription prices or model-provider usage charges.

How do you install Ruflo?

Ruflo can be initialized in a project folder with npx ruflo@latest init, followed by restarting Claude Code in that folder. Ruflo also offers Claude Code plugins, direct MCP client configuration and a standalone terminal console, with different capabilities across installation paths.

Does Ruflo work with Codex and other MCP clients?

Ruflo supports Codex and other MCP clients through its MCP server. Ruflo documents npx -y ruflo@latest mcp start as the connection command, allowing MCP clients to connect without the project initialization route.

Are Ruflo runtime mods sandboxed?

Ruflo runtime mods are not sandboxed and run with the user's account permissions, so the documentation advises reviewing their code before installation. Ruflo's console separately provides permission levels, approval settings, action logs and a take-back-control option.

From the community

No reviews yet

Be the first to review this tool. What you write helps the next person choose.

More tools like Ruflo

claude-code-harness

www.github.com

An open-source development harness for planning, implementing, checking, and reviewing AI-assisted coding tasks.

Free

Loom

www.github.com

An open-source delivery harness for planning, building, checking, and repairing work with coding agents.

Free

ECC

www.github.com

Open-source skills, agents, memory, and security tools for coordinated AI coding workflows.

Freemium

OpenRig

www.github.com

Build persistent AI coding agent teams with shared context, roles and coordinated work.

Free