Product tour

Inside the workflow workspace.

Edit a workflow, choose an MCP tool, review an AI edit, and inspect a run. These workbench examples use sample projects in Bandura 1.0.0; the download page lists the current release.

Bandura with the project files on the left, a checkout workflow on the canvas, and the selected request in the inspector Bandura with the project files on the left, a checkout workflow on the canvas, and the selected request in the inspector
The canvas shows the workflow. The inspector shows the selected step, and the bottom panel shows the run log. Development preview of the redesigned app.

Graph and code

Two views of the same process.

Edit the steps and routing visually, or work directly with the YAML in your repository.

The checkout workflow’s graph and YAML source open side by side in Split view The checkout workflow’s graph and YAML source open side by side in Split view
Split view keeps the graph and its .aether source together. Review the file in Git. Development preview of the redesigned app.

Read about graph and code editing

Results and variables

Inspect the values each step uses.

Open the resolved request and response, then inspect the values captured for later steps.

The node inspector displaying a resolved POST request with headers and a JSON body The node inspector displaying a resolved POST request with headers and a JSON body
Inspect the actual request produced by a step. Development preview of the redesigned app.
The response inspector showing HTTP 201 and the created order’s JSON body The response inspector showing HTTP 201 and the created order’s JSON body
Compare the sent request with the recorded response. Development preview of the redesigned app.

Explore the inspector panels →

MCP tools · 1.0.0

Choose the exact tool and its inputs.

Connect an MCP server, search its tools, and fill in nested arguments beside your workflow. Store the result in a variable for the next step.

MCP Configure inspector showing nested active and city filters and the customer output variable MCP Configure inspector showing nested active and city filters and the customer output variable
Authoring demonstration with a local fixture server and fictional customer data. Nested filters keep their boolean and string types, and the result is stored as customer. Searching tools does not execute them. Development preview of the redesigned app.

Connect and use MCP tools →

AI chat · 1.0.0

Review the file before it changes.

Choose which MCP connections chat can inspect. It can use exact tool schemas to propose a workflow, then show you the file for approval before saving.

Review new flow dialog showing the proposed customer lookup YAML and Discard and Apply controls Review new flow dialog showing the proposed customer lookup YAML and Discard and Apply controls
Scripted model demonstration of tool discovery and file review. This shows the approval flow, not a model-quality benchmark. Development preview of the redesigned app.

Build your first AI workflow →

Breakpoints

Pause before the next step runs.

Set a breakpoint, inspect the values available to the workflow, then step forward or continue. You can also select an earlier step to review its recorded result.

The checkout workflow paused before Create order, with Continue and Step controls, Create order selected, and variables open below The checkout workflow paused before Create order, with Continue and Step controls, Create order selected, and variables open below
This run is paused before Create order. The inspector and variables panel show the values Create order will use. Development preview of the redesigned app.

Follow the debugging guide

For API workflows

Keep requests aligned with your spec.

Compare a workflow with an OpenAPI definition and review suggested fixes when endpoints change.

Bandura’s OpenAPI drift dialog listing differences and proposed fixes for a workflow Bandura’s OpenAPI drift dialog listing differences and proposed fixes for a workflow
Review proposed fixes before applying them. The comparison does not require an AI model. Development preview of the redesigned app.

See how drift detection works →

Try a workflow in Bandura.

Start with the local order-process example. No account or AI key needed.

Download Bandura free