# Using Workflows and Tools > Trigger a specific tool or workflow from a local agent, the workflow page, the dashboard, the mobile app, or the API. Content type: documentation Source URL: https://www.agentpmt.com/docs/getting-started/using-workflows-and-tools Markdown URL: https://www.agentpmt.com/docs/getting-started/using-workflows-and-tools?format=agent-md Category: Getting Started --- # Using Workflows and Tools Every tool and workflow on AgentPMT is callable from four surfaces. Pick the one that fits how you are working. --- ## Tell a local agent exactly what to use Local agents (Claude Code, Cursor, Codex, Zed, anything connected over MCP) decide which tool to call based on what you say. Most of the time you can be loose: *"summarize this article."* When you want predictability, name the workflow or tool. ### By workflow name > Use the **Daily Market Digest** workflow with topic `AI infrastructure`. The agent looks for a workflow named "Daily Market Digest" in its MCP tool list and runs it. ### By tool name > Use the **Web Search** tool to find "MCP specification". Force a specific tool when the agent keeps picking the wrong one. ### By vendor > Use a Polygon.io tool to get the AAPL price. Useful when several tools could handle the same job and you want one specific provider. If the agent says it cannot find the workflow or tool, the most common cause is that it is not attached to your Agent Guardrail. Add it in **Control Center → Agent Guardrails**, then ask again — no reconnect required. --- ## Copy the prompt from the workflow page Every published workflow has a marketplace page with a one-click prompt copy. Use this when the agent is not reliably picking the workflow on its own, or when you want to paste into a chatbot that is not connected to AgentPMT. 1. **Open the workflow's marketplace page** From the marketplace, click any workflow. Its detail page opens. 2. **Click Use skill, then Copy prompt for LLM** The **Use skill** dropdown opens. Select **Copy prompt for LLM** ("Paste into your AI assistant"). The prompt copies to your clipboard with the workflow name, required inputs, and invocation instructions. 3. **Paste into your chatbot** Drop it into Claude, ChatGPT, Cursor, anywhere. Fill in the input fields the prompt asks for. Send. [Note title="Works even without an MCP connection"] The copied prompt is self-contained instructions for the chatbot. You do not need an MCP server up to use it — the prompt itself describes what to do. --- ## From the dashboard or the mobile app Hosted agents run tools and workflows on your behalf. To trigger a specific one: **Dashboard** — Open the agent in the Agent Builder, click **Open Chat** in the **Scoped chat preview** section, and type the workflow or tool name plainly: *"Run the Daily Market Digest workflow."* The agent calls it directly. **Mobile app** — Open the agent on your phone and type the same request. Activity logs the same way as on the dashboard; pending payment confirmations route to the device. The agent only runs tools and workflows that are attached to it (or to its Agent Guardrail). If a request does not match anything attached, the agent says so rather than guessing. --- ## Via the API Programmatic invocation uses your Agent Guardrail's **Bearer Token** in the standard `Authorization: Bearer ` header. The full request shape, streaming, polling long-running workflow sessions, error codes, and the response schema live in the dedicated [API tutorial](/docs/api-reference/programmatic-access). Endpoints are not duplicated here so they cannot drift. --- ## When something does not run | Symptom | Most likely cause | |---|---| | Agent says the workflow does not exist | Workflow is not attached to the Agent Guardrail or the agent. Add it under **Control Center → Agent Guardrails**. | | 401 from the API | Bearer Token is missing, expired, or copied with whitespace. Re-copy from **Show Bearer Token**. | | 402 from the API | Out of credits, or guardrail spending cap reached. Top up or raise the cap. | | Tool runs but errors out | Required credential not bound. Open the tool inside your Agent Guardrail and connect the credential. | --- ## Where to next - [Using Agents](/docs/getting-started/using-agents) - Run agents on every surface AgentPMT exposes. - [API Reference](/docs/api-reference/programmatic-access) - Full programmatic invocation reference.