Quick Start

This guide takes you from an empty account to a scheduled, audited workflow in five steps. You will connect a credential, build and test a workflow, create a Hosted Agent to run it, put it on a schedule, and audit the run.

Each step builds on the last, so do them in order the first time through. If you already have some of these pieces in place, skip ahead.

Connect a credential

A credential is the secret your agent uses to call an external service — an API key, an OAuth token, or similar. Not every tool needs one, but any tool that hits a service requiring your own account will. You create the credential once, then bind it to the Agent Group that will run the workflow.

1. Open the Credentials tab

   In the dashboard, go to the Credentials tab and click Add.

   

2. Choose the credential type and fill in the fields

   Pick the credential type the tool expects (for example, an API key). Secret fields are masked after you save and are never displayed back. See Credentials for what each type means.

   

3. Bind the credential to your Agent Group

   In your Agent Group settings, bind the credential to the tool that needs it. AgentPMT injects the secret server-side at call time — your agent never sees the raw value.

   

   

Build and Test a Workflow

A workflow chains several tools into a reusable recipe that runs in one go — for example, "search the web, scrape the top results, summarize, and email me." You build it once, then attach it to an agent or call it directly.

1. Open the Workflow Builder

   Go to the Workflow Builder. It opens in Workflow mode with a blank canvas.

2. Name it and describe the outcome

   Give the workflow a name and a one-sentence description of what it should produce. The description is what an agent reads to decide when to use the workflow, so be concrete.

3. Drop in tool nodes and connect them

   From the sidebar, search the marketplace and click a tool to add it as a node. Drag a line from one node's output to the next node's input to set both the execution order and the data flow.

4. Configure each node's inputs

   Click a node to open its config panel. Each input can be a literal value, a reference to an earlier node's output, or a workflow input that gets filled in at run time.

5. Test it from the canvas

   Run the workflow from the canvas. Each node lights up as it executes, and errors surface on the node that produced them. Test tools that need a credential only after you have bound it to the Agent Group.

6. Publish or keep private

   Click Publish to share it to the marketplace, or keep it private. Either way it is now available to attach to an agent. See Build a Workflow for the full walkthrough.

Create a Hosted Agent

A Hosted Agent runs inside AgentPMT. You build it once, then use it from the dashboard, the website, the mobile app, or via API. It decides on its own which tool or workflow to use based on your message.

1. Open the Agent Builder

   Go to the Agent Builder, or open the marketplace and click Build custom agent at the top.

2. Name your agent and write a one-line description

   The description tells the agent how to behave. Be concrete: "Runs the daily research workflow and emails me the summary" beats "research helper."

3. Attach your workflow (and any tools)

   From the canvas Catalog, click Add workflow on the workflow you just built. Add any individual tools you want the agent to call directly.

4. Try it in the scoped chat

   Scroll to the Scoped chat preview section and click Open Chat. Ask the agent to do something that uses your workflow. You should see the tool calls appear in the conversation and a corresponding entry in Control Center → Agent Activity.

Run On a Schedule

Instead of triggering the workflow by hand each time, put it on a schedule and AgentPMT runs it on a recurring basis — daily digests, hourly checks, weekly reports.

1. Open Schedules

   Go to Control Center → Workflows → Schedules.

2. Create a schedule for your workflow

   The schedule attaches to the Agent Group that runs the workflow, so it can only target workflows that group can access — workflows you added explicitly, public workflows, or private workflows granted to the group.

3. Choose how it runs

   With an AgentPMT managed run, AgentPMT's own runner executes the workflow on schedule using the model assigned to it. (You can also bring your own scheduler that claims due items from a queue — see Using Workflows and Tools for that path.)

4. Test the schedule

   Use Test Run to fire the workflow once with the stored prompt without advancing the next due time. Confirm it produces the output you expect.

Audit the Run

Every call your agent and workflows make is recorded. Use the activity log to audit a specific run, confirm whether it consumed credits, or export usage for review.

1. Open Agent Activity

   Go to Control Center → Agent Activity. Each row is one tool call with its cost, duration, and status.

2. Filter to the run you care about

   Filter by Agent Group, time window, or status to narrow down to the scheduled run you want to inspect.

3. Expand a row for the full request and response

   Expand any row to see the full request and response. This is where you confirm the workflow did what you expected and spot any failures.

4. Export if needed

   Click Export to CSV to pull the filtered activity out for review or record-keeping.

Where to next