Vercel Adds Agent Runs Tracing Capability to MCP Server and CLI
Vercel has expanded its Agent Runs feature, allowing developers to retrieve full execution traces, reasoning steps, and token usage of AI agents directly via the Vercel MCP server and Vercel CLI.
Vercel has extended its Agent Runs functionality, built for the open-source eve AI agent framework, to be directly accessible from the Vercel MCP server and the Vercel CLI. This integration automatically captures execution logs of agents deployed on Vercel. Developers can now inspect complex agent execution details directly inside their editors or terminals without needing to open a web browser.
Primary source
Open Vercel Changelog ↗This article summarizes the source. Check the original for full details and conditions.
Comparison
| Aspect | Before / Alternative | After / This |
|---|---|---|
| Trace Inspection Access | Limited to the Vercel web browser interface | Available directly in-editor via MCP or in terminal via Vercel CLI |
| Command Availability | Manual dashboard exploration only | Dedicated commands like list_agent_runs and get_agent_run |
| Trace Detail Level | Standard application logs and request metrics | Deep reasoning steps, tool calls, sub-agent behaviors, and token usage |
Action Checklist
- Update the Vercel CLI to the latest version Ensure your local environment has the required CLI updates for agent commands.
- Configure the Vercel MCP server in your editor This enables IDEs supporting the Model Context Protocol to fetch run details.
- Deploy your eve-based agent framework project to Vercel Agent Runs specifically tracks AI agents constructed with the open-source eve framework.
- Query agent traces using the CLI or MCP commands Use list_agent_runs or get_agent_run to retrieve model reasoning and token usage.