# AI use cases for OpenFlow (SeeRay11/OpenFlow)

Repo: [SeeRay11/OpenFlow](https://github.com/SeeRay11/OpenFlow) | Use cases: 4 | Page: https://dedupedai.com/use-cases/seeray11-openflow/

OpenFlow is a visual canvas for building multi-agent AI workflows, built as a fork of the opencode engine, that runs agents which read and write real files in a chosen repo.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Run a planner, architect and coder pipeline on a repo: https://dedupedai.com/use-cases/seeray11-openflow/run-planner-architect-coder-pipeline/
- Try a multi-agent run with no API key: https://dedupedai.com/use-cases/seeray11-openflow/try-multi-agent-runs-with-free-models/
- Connect paid provider keys for agent runs: https://dedupedai.com/use-cases/seeray11-openflow/connect-paid-provider-keys-for-agent-runs/
- Review each agent permission before it edits files: https://dedupedai.com/use-cases/seeray11-openflow/review-agent-permissions-before-edits/

## How to run a planner, architect and coder pipeline on a repo with OpenFlow

Repo: [SeeRay11/OpenFlow](https://github.com/SeeRay11/OpenFlow) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Wire role cards into a saved pipeline and run parallel agents against a local codebase.

**Who it is for:** A developer who wants several AI agents to split up a coding task, such as planning, designing and implementing a change, instead of prompting one agent at a time.

**The problem:** A single agent prompt tends to mix planning, design and coding in one pass, and there is no visible structure to reuse or inspect.

**The outcome:** A saved, named pipeline of role nodes that runs agents in parallel against the target repo, with each step visible on the canvas.

### Steps

1. Install Bun 1.3 or newer and Git, then check Bun with `bun --version`.
2. Run `git clone https://github.com/SeeRay11/OpenFlow.git`, then `cd OpenFlow`, then `bun install`.
3. Set the OPENFLOW_PROJECT environment variable to the path of the repo the agents should work in. These agents edit real files.
4. Run `FLOW_MANAGE_SERVER=1 bun openflow.ts` (PowerShell: `$env:FLOW_MANAGE_SERVER=1; bun openflow.ts`), then open http://localhost:5174.
5. Drag role cards onto the canvas, wire them as planner to architect to coder, save the pipeline, and run it.

### Prompt to paste into your agent

```text
Using the OpenFlow repo at https://github.com/SeeRay11/OpenFlow, set OPENFLOW_PROJECT to <target-repo-path>, start it with `FLOW_MANAGE_SERVER=1 bun openflow.ts`, and build a saved planner, architect and coder pipeline on the canvas at http://localhost:5174. Run it and report the final run log. Verify that each node ran as its own agent and that the run log lists a decision for each permission request.
```

### Caveat

The agents write real files in OPENFLOW_PROJECT, which defaults to the OpenFlow repo itself, so point it at a repo you are willing to change.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/seeray11-openflow/run-planner-architect-coder-pipeline/

## How to try a multi-agent run with no API key with OpenFlow

Repo: [SeeRay11/OpenFlow](https://github.com/SeeRay11/OpenFlow) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Use the free -free model tier to test a pipeline before connecting any paid provider.

**Who it is for:** A developer evaluating OpenFlow who wants to see a real agent run without setting up billing or keys.

**The problem:** Setting up provider keys and billing before knowing whether a visual agent pipeline fits a workflow is a costly first step.

**The outcome:** A completed trial run on free-tier models, plus a clear sense of whether the pipeline style works for the team.

### Steps

1. Install Bun 1.3 or newer and Git, then run `git clone https://github.com/SeeRay11/OpenFlow.git`, `cd OpenFlow`, and `bun install`.
2. Run `FLOW_MANAGE_SERVER=1 bun openflow.ts` (PowerShell: `$env:FLOW_MANAGE_SERVER=1; bun openflow.ts`) and open http://localhost:5174.
3. On each node, pick a model whose id ends in `-free`, then use the test button beside the model before starting a full run.
4. If a run returns a 429 or a 400 from the model, pick another `-free` model, because the shared quota is busy and the install is not broken.

### Prompt to paste into your agent

```text
Using the OpenFlow repo at https://github.com/SeeRay11/OpenFlow, start the canvas with `FLOW_MANAGE_SERVER=1 bun openflow.ts`, assign only `-free` models to each node of a planner, architect and coder pipeline, and test each model with the test button before running. Report which models answered and which returned errors. Verify that the run finishes or that each failure is a 429 or 400 from a busy free model.
```

### Caveat

The free tier is a shared quota, so a busy model can fail even when the install is correct.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/seeray11-openflow/try-multi-agent-runs-with-free-models/

## How to connect paid provider keys for agent runs with OpenFlow

Repo: [SeeRay11/OpenFlow](https://github.com/SeeRay11/OpenFlow) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README

Add provider API keys in the titlebar so paid models can run pipeline nodes.

**Who it is for:** A team lead who has provider accounts and needs paid models for larger pipeline runs than the free tier allows.

**The problem:** Paid models will not run until a key is connected, and keys already held by the opencode CLI are not visible to the server until imported.

**The outcome:** Paid models available to nodes, with keys taking effect without a server restart.

### Steps

1. Start OpenFlow with `FLOW_MANAGE_SERVER=1 bun openflow.ts` and open http://localhost:5174.
2. Click "api keys" in the titlebar, pick a provider, and paste its key in the two-step connect dialog.
3. If keys already exist in the opencode CLI, run `opencode providers login` and import them from the same panel, because the server does not read auth.json for its model catalog otherwise.
4. Use the test button beside a paid model on a node to confirm it answers before a full run.

### Prompt to paste into your agent

```text
Using the OpenFlow repo at https://github.com/SeeRay11/OpenFlow, start the canvas, connect a provider key through the "api keys" panel, and assign one paid model to a node in a planner, architect and coder pipeline. Test the model with the test button and report the provider used. Verify that the model appears as available to the node without restarting the server.
```

### Caveat

A paid model needs a connected key, and a fresh install with no key cannot spend money.

Risk, credentials: Provider API keys pasted or imported here give the agents paid access, so use keys scoped to this work and do not share the canvas session.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/seeray11-openflow/connect-paid-provider-keys-for-agent-runs/

## How to review each agent permission before it edits files with OpenFlow

Repo: [SeeRay11/OpenFlow](https://github.com/SeeRay11/OpenFlow) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README

Switch to ask-me mode and read the run log to approve each agent action.

**Who it is for:** A cautious engineer who wants to see and approve each file change an agent proposes before a pipeline is trusted on a real repo.

**The problem:** Auto-approved permissions let agents edit real files without a pause, which is hard to trust on an unfamiliar codebase.

**The outcome:** A pipeline run where each permission request is reviewed by a person, with every decision recorded in the run log.

### Steps

1. Start OpenFlow with `FLOW_MANAGE_SERVER=1 bun openflow.ts` and open http://localhost:5174.
2. Set the permissions toggle in the toolbar to `ask me` instead of the default `auto`.
3. Run the saved pipeline and approve or deny each permission request as it appears.
4. Open the run log and check that every decision is recorded for the run.

### Prompt to paste into your agent

```text
Using the OpenFlow repo at https://github.com/SeeRay11/OpenFlow, set the permissions toolbar to ask me, run a saved pipeline against <target-repo-path>, and list each permission request with its decision. Output a summary of the file changes approved. Verify that the run log contains one entry per decision.
```

### Caveat

Permissions default to auto, which approves each request for that one call, so ask me mode must be selected on purpose before a run.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/seeray11-openflow/review-agent-permissions-before-edits/