Skip to main content
Glama
Open Flow MCP Server

Open Flow MCP Server

Official
by oomol-lab

English | 简体中文 | 繁體中文 | 日本語 | 한국어 | Русский | Français

CI npm License: Open Flow Node.js 26 Bun 1.4

Providers Actions

Build automation with your AI agent. See exactly what will run.

Open Flow is a source-available workflow automation platform for teams using coding agents. Build and edit workflows with Codex, Claude Code, or another coding agent through the CLI or MCP. Review and test the same workflows in Workbench before publishing.

Combine JavaScript, connected apps, AI, branching, and human approval in one canvas. Inspect the inputs, follow the execution, and decide which version goes live. Run on OOMOL Hosted or on your own infrastructure.

Explore the interactive demo · Use OOMOL Hosted · Self-host with Docker

A vertical AI trend workflow in the Open Flow Workbench, with parallel X, GitHub, and Hugging Face sources and the Research & write node's prompt, tools, inputs, and output in the property panel

Your agent builds. You stay in control.

Describe the work you need done:

“Read new support emails, classify each request, draft a reply, and ask me to approve it before sending.”

Through oo flow, a terminal agent can discover integrations, create and edit nodes, check a draft, run it, inspect results, and publish when you ask. Its changes appear in the same Workbench you use to review the graph and edit the code. There is no separate AI-generated project to translate or keep in sync.

The Server also exposes MCP authoring and run tools for compatible clients. Both interfaces work against the selected deployment's saved Flows, revisions, and runs.

Use an agent for the repetitive wiring. Keep the decisions that matter visible.

Related MCP server: Heym

Put every step where you can inspect it

A workflow should still make sense after the conversation that created it is gone.

  • JavaScript for precise work. Transform data, format messages, or write custom logic in Code Tasks with named, typed inputs and outputs.

  • Explicit input sources. See where a value comes from in the property panel. Control connections determine what executes next; input mappings determine what data a step receives.

  • Branches and reusable subflows. Route work with conditions and give repeated logic a clear interface of its own.

  • Notes alongside the work. Explain a decision directly on the canvas so the next person can understand the intent.

The current Code Task property panel showing JavaScript, a customer input source, and a typed message output

Give AI a job with boundaries

Use LLM Tasks to classify, extract, or summarize. Use Agent Tasks when a job needs several tool calls. Surround them with ordinary code, conditions, and approvals where you need predictable behavior.

An Agent Task has a configured model, declared tools, and execution limits. Its tool configuration specifies the actions, accounts, and approval policy available to the model. You can require human approval for a tool call, or place an Approval node in the workflow before the next step proceeds.

Approval and Wait nodes persist their waiting state. Work can resume after a decision without replaying the steps already completed in that run.

See what happened. Choose what goes live.

Test a draft from a chosen trigger and follow its execution in the Workbench. Inspect node results, logs, errors, and pending approvals instead of reconstructing a run from the final message alone. Run history keeps the execution tied to the revision that produced it.

An onboarding test run paused for human approval in the Open Flow Workbench

Publishing creates a versioned snapshot for Live automation. Continue editing the draft while the published version runs, inspect earlier Publications, and roll back when needed.

Start work manually, on a schedule, through a webhook, or from supported provider events and polling sources. The same deployment owns the workflow, its Live version, and its execution state.

Connect apps without putting credentials in the graph

Open Flow uses a Connector runtime such as OpenConnector to discover and execute actions for services including Gmail, Slack, GitHub, and Notion. Account credentials stay with the Connector; workflows refer to Connection identities.

OOMOL Hosted provides managed OAuth apps for supported integrations. With a self-hosted stack, you choose the Connector deployment and manage the required provider configuration and account authorizations. Workflow logic and account access remain separate concerns.

See Open Flow with OpenConnector and the oo CLI for a complete setup, including authorizing an account and building a first Flow.

Run it your way

Option

What you manage

Get started

OOMOL Hosted

Your workflows and connected accounts; OOMOL operates the deployment.

Open OOMOL

Docker

Your deployment, storage, backups, upgrades, and integrations.

Commands below

Fly.io

Your app, persistent volume, secrets, backups, and upgrades on Fly's infrastructure.

Deployment guide

For a local self-hosted instance, install Docker and OpenSSL, then run:

git clone https://github.com/oomol-lab/open-flow.git
cd open-flow

export OPEN_FLOW_TOKEN="$(openssl rand -hex 32)"
docker build --file apps/server/Dockerfile --tag open-flow-server:dev .
docker run --rm \
  --publish 3000:3000 \
  --env OPEN_FLOW_TOKEN="$OPEN_FLOW_TOKEN" \
  --volume open-flow-data:/data/open-flow \
  open-flow-server:dev

Open http://127.0.0.1:3000 and sign in with OPEN_FLOW_TOKEN. Flows and run history are stored in the Docker volume. The operator token also authenticates CLI and API access.

Prefer a prebuilt image? Follow the GHCR image guide and select an explicit release tag. The latest tag is reserved for stable releases.

Connector actions and LLM Tasks need their corresponding services configured. The Server does not silently fall back to another provider. Before exposing a deployment publicly, follow the deployment guide and hardening checklist.

Build on Open Flow

Open Flow uses the Open Flow License, based on Apache-2.0 with additional conditions. The repository includes the workflow contracts and runtime, Workbench, CLI command package, and self-hosted Server. The versioned Control API keeps clients independent of the deployment's storage and execution implementation.

Use the versions of Bun and Node.js pinned in .bun-version and .node-version:

bun install --frozen-lockfile
bun run dev

The development Workbench opens at http://localhost:5174. The first run creates an operator token in apps/server/.open-flow-dev/operator-token. See CONTRIBUTING.md for configuration, checks, and the component Lab.

Learn more

Reference

Product model and runtime boundaries

Architecture

Integrate your own clients

Control API · MCP

Run and operate a deployment

Server · Docker · Fly.io

Connect apps and an AI coding agent

OpenConnector + oo CLI

Browse all documentation

Documentation index

  • OpenConnector: source-available connector gateway that provides the Provider catalog, credentials, and Action execution behind Connector-backed nodes.

  • oo CLI: local agent toolkit that hosts the oo flow command built from this repository.

Contributing

Issues and pull requests are welcome. Read CONTRIBUTING.md for the development setup, repository rules, and checks to run before opening a pull request. Participation in this project is governed by CODE_OF_CONDUCT.md.

Security

Please report vulnerabilities privately through GitHub private vulnerability reporting rather than public issues. SECURITY.md describes the supported versions, the disclosure process, what is in scope, and how to harden a self-hosted deployment.

License

Open Flow License. Third-party notices for bundled assets are listed in NOTICE.

Contributors

Thanks to everyone who has helped build Open Flow. Want to join them? See CONTRIBUTING.md.

Open Flow contributors

Star History

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      C
      quality
      A
      maintenance
      Self-hosted, source-available AI workflow automation platform. Build multi-agent, RAG, and tool-using pipelines on a visual canvas and publish any workflow as an MCP server (stdio/SSE/Streamable HTTP). Also an MCP client via the agent node.
      2
      1,820
      -
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables plain-English workflow automation across connected tools like GitHub, Notion, Slack, and more, with self-healing execution, scheduling, and status reporting through MCP.
      15 npm
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      An agent workflow engine exposed over MCP that lets AI agents execute node-graph workflows, following per-step directives with explicit completion conditions and JSON-Schema-validated inputs and outputs to move between success, failure, and conditional branches. It can be self-hosted in a single Docker container with OAuth 2.1 authentication and a supplementary web UI for managing and visualizing workflows.
      126
      Apache 2.0