BANDURA Download

Free local AI workflow builder

Build AI workflows.
Run them locally.

Connect your AI model, APIs, and scripts on a visual canvas. Check every result, debug any step, and run the workflow again with new inputs.

Download Bandura See the product tour

Free for personal and commercial use. No account required.

Start with a local sample, no AI key needed. Add your own model when you’re ready.

macOS, Windows and Linux · v1.0.0
Unsigned macOS builds: installation guide

The desktop workspaceTake a closer look
Bandura displaying a checkout workflow, the selected request and response, and the run console Bandura displaying a checkout workflow, the selected request and response, and the run console
Follow the workflow on the canvas. Select any step to inspect the request, response, and variables. Development preview of the redesigned app.
  • Local execution
  • Editable YAML
  • Git versioning
  • Bring your own model

Start with a real task

A workflow you can download.

Try classification, JSON extraction, or a labelled evaluation. Complete examples include setup instructions and synthetic data.

Browse all workflows, including an API example with no AI key →

Watch a workflow pass, then fail a check.

A recorded walkthrough of the development build: a synthetic message, a deterministic local response, and an assertion you can inspect. This demonstrates workflow behavior, not model accuracy.

Read the walkthrough
  1. Open the support classification flow with the synthetic message “I was charged twice for my order.”
  2. Run the workflow. The demonstration server returns billing; the allowed-category check passes.
  3. Select the model step and inspect its recorded result.
  4. Open a copy whose assertion deliberately expects an impossible category. Run it again.
  5. The assertion fails. Inspect that step and its variables to understand the mismatch.

Download the workflow and use your own model →

Use cases

AI workflows
and API tests.

Start with a process you already know. Put the inputs, model calls, and checks in one workflow.

02

Use a local model for private work

Connect Ollama or LM Studio to run AI steps on your hardware. Your APIs and connected tools still receive the calls you send them.

Explore local model setup
03

Bring your API tests with you

Import existing collections, chain requests, and check the results. Keep readable workflow files alongside your code.

Compare Bandura with Postman

Visual editor and YAML

Two views.
The same workflow.

Connect steps on the graph or edit the YAML. Both views update the same .aether file, so every change can live in your repository.

The YAML source of a Bandura workflow open in the code editor The YAML source of a Bandura workflow open in the code editor
Readable workflow files you can diff, commit, and share with your team. Development preview of the redesigned app.
  1. 01

    Connect the steps

    Add model calls, HTTP requests, scripts, and branches. Pass each result to the step that needs it.

  2. 02

    Check what comes back

    Set assertions for the output you expect. Try sample inputs from a CSV or JSON dataset.

  3. 03

    Run it again. Or hand it over.

    Reuse the file with new inputs. Share the project and its dependencies; your teammate adds their own credentials.

Requests and results

See what
each step sent.

Inspect the resolved request, response, and variables. Set a breakpoint, step through execution, or open a previous run from local history.

See the debugger in the product tour
The selected step’s resolved POST request, URL, status, and JSON body in Bandura’s inspector The selected step’s resolved POST request, URL, status, and JSON body in Bandura’s inspector
The request after variables have been resolved, including the body sent to the API. Development preview of the redesigned app.

Check the request behind the result

Confirm the method, URL, and body a step used. Switch to its result to inspect the response, or open the variables panel to trace captured values.

Run history keeps recorded results available for review without sending the request again.

Read the inspector guide

Your model, your account

Use Anthropic, Gemini, an OpenAI-compatible endpoint, or a local model through Ollama or LM Studio.

Connect a model

APIs and MCP tools

Call HTTP APIs or add tools from a remote MCP server. Import existing Postman, OpenAPI, or curl requests.

Connect your tools

From desktop to terminal

Run workflows with the CLI in your terminal or CI. Expose them to a compatible AI assistant through the local MCP server.

Use the CLI

Getting started

Your first workflow,
in three steps.

You don’t need a model account to try Bandura. Run a local sample first, then add AI and connected tools to a workflow you understand.

  1. 01

    Install the desktop app

    Choose your platform. No signup or payment. Early-access installation notes are next to the downloads.

    Get the free app
  2. 02

    Run a local sample

    Follow the order-process example: calculate a total, check it, then change the expected result to see a failure.

    Open the starter guide
  3. 03

    Add your model

    Connect a hosted provider or local endpoint. Try labelled sample inputs before using real customer data.

    Connect your AI model

Free, including at work

The whole app.
No subscription.

Build personal projects or workflows for your team. There are no paid seats, activation keys, or feature tiers.

You supply your own model and service accounts. Their usage charges still apply.

Support Bandura on Patreon
DESKTOP APP

Install Bandura

macOS, Windows and Linux

  • Free for personal and commercial use
  • No account required
  • Workflows stored on your machine
Download Bandura

Stable v1.0.0 · Installation notes

Before you start

A few useful answers.

Need help with your setup?
[email protected]

Is Bandura free?

Yes, for personal and commercial use. There are no paid seats or feature tiers. Optional Patreon support funds development and unlocks no extra features.

Which AI models can I use?

Anthropic, Google Gemini, or any OpenAI-compatible endpoint, including local models through Ollama and LM Studio. You use your own account and pay any provider usage charges directly.

Set up your model

Does my data leave my computer?

Workflows and run history are stored locally. Remote model providers and APIs receive the data you send them. You can use a local model; the optional update check can be disabled. The app has no telemetry or crash reporting.

Read the data-flow details

Does Bandura run on a schedule or in the cloud?

The desktop app runs workflows locally when you start them. Built-in scheduling and hosted execution are not available. You can invoke the CLI from your own scheduler or CI environment.

Can I use it for API testing?

Yes. Chain requests, pass values between steps, and assert the results. Import Postman, Insomnia, Bruno, OpenAPI, HAR, or curl to bring existing API work into Bandura.

Import your requests

Can I share workflows with my team?

Yes. Share the YAML file or commit the project to Git, including its scripts, subflows, and datasets. Each person supplies their own credentials and model settings. Bandura is proprietary software; your workflow files remain yours.

Stable release · v1.0.0New features, fixes, and installation details.

Read the release notes

Try Bandura

Run your first workflow.

Download the free app and run a local sample. No signup, payment, or AI key needed to start.

Download Bandura free

Read the step-by-step starter guide