Running Agents
Markup AI’s content agents are exposed through the Agents endpoints of the API. Each agent runs against the settings configured for your organization in the Markup AI console — no per-call style guide IDs. For what each agent does and why it matters, see the Agent Catalog.
The run flow
- Discover the agents available to your organization with
GET /agents. Each entry includes the agent’s stablename(e.g.brand_voice), itsid, description, and category. Resolve theidfrom thenameat runtime — every agent page below has copy-paste examples that do exactly that. - Run an agent with
POST /agents/{agent_id}/run, passing your content astext. Passwait=trueto block until the run completes; otherwise the response returns immediately with aworkflow_id. - Track longer runs with
GET /agents/workflows/{workflow_id}, stream progress fromGET /agents/workflows/{workflow_id}/stream(server-sent events), or supply awebhook_urlin the run request to receive the result when the agent finishes.
The agents returned by GET /agents depend on your organization’s plan and feature access.
Agent reference
Via MCP
The same agents run on the hosted MCP server at https://api.markup.ai/mcp/, addressed by the same name used here. Ask for a review and markupai_review picks the agents for you, or name them explicitly with markupai_run_agents — see the MCP overview to connect Claude, Cursor, or any other MCP client.