Pipeline Builder: Chain Tools into Reusable Workflows

A repeated task often has more than one step: resize an image, convert its format, or scrape a public page and transform the resulting text. Agent Media Tools’ Pipeline Builder lets signed-in users save those sequences and run them consistently from the Toolbox Workspace.

What the shipped builder supports

Steps execute sequentially. Each successful output is merged into the working data for the next step. A step can also include a simple condition—equals, contains, or not-empty—and an on_error policy.

Start with the browser

1. Sign in and open the Pipeline Builder.
2. Choose a template or create a blank pipeline.
3. Add steps, configure their parameters, and save.
4. Run the pipeline and inspect the result for every step.

The current templates cover Image to WebP, Scrape & Slugify, JSON Beautifier + Hash, Text Cleaner, and Image Optimizer. Templates are starting points: cloning one creates an editable pipeline under your account.

The session-authenticated REST API

The same human account session used by the Toolbox can manage pipelines through these routes:

These management and execution routes require a signed-in browser session. For scripts, establish a session through the login endpoint and keep the returned cookie. They are not advertised as API-key or MCP pipeline routes.

Create a text workflow with curl

💡 Windows users: Run these commands from Git Bash or WSL. PowerShell aliases curl to Invoke-WebRequest.

curl -c cookies.txt \
  -H "Content-Type: application/json" \
  -d '{"username":"YOUR_USERNAME","password":"YOUR_PASSWORD"}' \
  https://agentmediatools.com/api/login

curl -b cookies.txt \
  -H "Content-Type: application/json" \
  -d '{
    "name":"Clean and hash text",
    "description":"Trim text, create a slug, then hash it",
    "config":{"steps":[
      {"id":"trim","name":"Trim","tool":"text-transform","params":{"action":"trim"}},
      {"id":"slug","name":"Slugify","tool":"text-transform","params":{"action":"slugify"}},
      {"id":"hash","name":"SHA-256","tool":"hash-calculate","params":{"algorithm":"sha256"}}
    ]}
  }' \
  https://agentmediatools.com/api/pipelines

The create response contains the pipeline ID. Use it with the owned execution route:

curl -b cookies.txt \
  -H "Content-Type: application/json" \
  -d '{"input":{"text":"  Product Launch Notes  "}}' \
  https://agentmediatools.com/api/pipelines/YOUR_PIPELINE_ID/run

The response includes a run ID, overall status, ordered step results, and the merged final output. Runs are synchronous: the HTTP request remains open until the steps complete or fail.

Conditions and failure behavior

A condition controls whether a step runs. This example hashes text only when the working text field is not empty:

{
  "id":"hash",
  "name":"Hash non-empty text",
  "tool":"hash-calculate",
  "params":{"algorithm":"sha256"},
  "condition":{"type":"not_empty","field":"text"},
  "on_error":"stop"
}

With on_error: "stop", an error stops the remaining sequence and marks the run failed. Without it, the engine records the failed step and continues, so choose the policy deliberately.

Pipeline webhooks

Runs started through the owned session route can emit pipeline.completed or pipeline.failed to active account webhooks subscribed to those events. The payload contains the pipeline and run IDs, pipeline name, status, step count, duration, and a bounded error field—not the full run result.

Webhook delivery failures do not change a completed pipeline response. Use Webhooks to configure destinations, and use the run-history route when you need detailed step results.

Practical patterns

• Image optimization: resize an existing server-side file, then convert it to WebP or JPEG.
• Public-page normalization: scrape a public URL, then trim or slugify the extracted text.
• Deterministic text processing: normalize text and calculate a repeatable hash.
• JSON inspection: format current working data before hashing or reviewing it.