Suede AI Agent StudioDocsStart buildingOpen workspace
Navigation menu

Agents & workflows5 min readUpdated

Build your first AI agent in Suede AI Agent Studio

Quick answer

To build your first agent in Agent Studio, open Guided, describe the job in one sentence and answer a few setup questions. Agent Studio drafts the workflow, lists every step for your review, and opens it on the canvas, where a dry-run test costs nothing. Save a version, then launch it and let it run on its own.

What you'll have

  • A working agent drafted from one sentence, with every step visible before anything runs
  • A free dry-run test that proves the wiring, the branches and the output shape
  • A saved version ready to launch the moment it looks right
On this page
  1. What do you need before you build an agent?
  2. How do I build an AI agent step by step?
  3. Should I start from a template or a blank canvas?
  4. What does a working agent look like inside?
  5. What does it cost to build and test an agent?
  6. Questions

What do you need before you build an agent?#

You need one job that keeps coming back. Lead follow-up, overdue invoices, meeting prep and support triage are the classic first picks because they repeat every week and follow the same pattern each time. Write the job down the way you would brief a new hire: what comes in, what a good result looks like, and where it should land.

Not sure which job to hand off first? Put a number on the ones you have. The calculator below shows what a recurring task costs you in hours and dollars each month. The biggest number is your first agent.

Your numbers

What one repeating task costs you

Pick a task you do every week: follow-ups, reports, data entry. Replace the sample numbers with yours.

Hours back each month

10.8

Worth each year

$6,500

That is 130 hours a year on one task. Describe it once and an agent runs it on schedule from then on.

How do I build an AI agent step by step?#

Guided turns a plain sentence into a wired workflow. The whole path runs through three stages on screen: Describe the job, Review the workflow, and Choose what happens next.

Your first agent, start to launch

0 of 6 done

  1. Step 1: Describe the job

    Choose Describe it and type the outcome you want, for example Text me the moment a hot lead comes in. Press Start guided build.
    Open Guided
  2. Step 2: Answer the setup questions

    Agent Studio asks what it needs to know, one question at a time. Reply and press Send answer. Workspaces with credit get an interview that reads your answers and asks only what your particular workflow needs.
  3. Step 3: Review the workflow

    The review screen lists The steps in order, with the tool each one uses. Fill in the Job brief and an Expected output so the goal travels with the agent. Want something different? Press Change anything and say what to adjust.
  4. Step 4: Open it on the canvas

    Press Open on canvas. Agent Studio saves the draft and opens it in Studio with every step wired, so you can connect your tools and fine-tune prompts.
  5. Step 5: Run a free test

    Press Run test. Every step streams its result into the run dock and nothing spends credit. Fix anything that looks off and run it again. See Test your agent free for what a dry run proves.
  6. Step 6: Save a version and launch

    Open Versions and press Save version to freeze a checkpoint, then launch. Your agent shows up on your dashboard with its own page and starts working on its trigger.
    Open your agents

Should I start from a template or a blank canvas?#

Start from whichever door matches what you already know. The start page offers all three side by side.

Three ways to start an agent
Start withPick it whenWhere
Describe itYou know the job and want Agent Studio to work out the stepsGuided
Start from a templateA ready-made agent already does most of the jobTemplates
Blank canvasYou want to place and wire every step yourselfStudio

What does a working agent look like inside?#

Every agent is a flow of steps called nodes, joined by connections that carry data from one step to the next. The smallest useful agent has three: Input takes what comes in, LLM (Claude) does the thinking, and Output returns the result. Real agents add more:

  • A trigger, such as Schedule, Webhook or CRM Trigger, so the agent starts itself. See Triggers and schedules.
  • AI steps such as Classify and Extract Fields that sort and pull structure out of messy text.
  • Logic such as Branch, Switch and Loop that route each case down the right path.
  • Action steps such as Send Text, Send Email, Create Task and Slack Message that do the work where your team already lives.

The full catalog, with what each node does and costs, is in the node reference.

  • 3 ways inDescribe it, Start from a template, or Blank canvas
  • 4 stagesCreate, Configure, Test and Launch, shown on the canvas as you go
  • $0to build, edit, save versions and run dry-run tests

What does it cost to build and test an agent?#

Building costs nothing. Drafting, editing, saving versions and dry-run tests are all free. Live runs draw from your prepaid credit, so you pay for real work your agent does and nothing for the hours you spend getting it right. Current rates are on the pricing page.

Questions buyers ask#

Do I need to know how to code to build an AI agent?

No. Guided builds the agent from a plain-English description and a few questions, and Studio lets you adjust it by dragging and connecting steps. Code is there for developers who want TypeScript, and nobody else ever has to open it.

How long does it take to build my first agent?

One conversation in Guided produces a wired draft, and a template opens fully built. Most of your time goes into reviewing the steps and running a test or two, which is exactly where it should go.

Is testing my agent free?

Yes. Dry-run tests are free: logic and data steps run for real, while model calls and outside tools return simulated results. When you want real model output, turn on Run for real, which uses credit.

Can I build an agent on my phone?

Yes. Guided works on your phone, so you can describe the job, answer the questions, review the steps and save the draft from anywhere. Open Studio on a laptop or desktop to run tests and wire connections.

What kind of work can my first agent do?

Anything that follows a pattern: scoring leads, drafting invoice reminders, writing meeting briefs, triaging support tickets, replying to reviews or running a weekly SEO audit. Agent Studio ships more than 100 templates for these jobs across Sales, Marketing, Finance, Support, Ops, Engineering, HR and Legal.

How is an Agent Studio agent different from a GoHighLevel workflow?

A GoHighLevel workflow chains triggers and actions on CRM records. Agent Studio covers that with its CRM Trigger and CRM action steps, and adds AI steps that read, classify and write, plus a free dry-run test and numbered versions before anything goes live.

Was this page helpful?

Good work starts with a clear flow.

Give an agent a job. Connect the steps. See how the work becomes a service someone else can call.

Input
Reason
Output

A workflow you can inspect

Make the steps visible.

Start with a template or describe the job. Connect inputs, reasoning, and output on the canvas.

Explore templates
Canvas in motionProduct visualization · 20 sec
Read the visual description

A stylized organization chart opens into a workflow canvas. Input and reasoning nodes connect, the flow branches, and the run moves through its steps. The film ends with Suede AI Agent Studio. Music only; no spoken narration.