MCP Server
Quickstart

Quickstart

Get the Tenderly MCP Server running in your AI assistant in under a minute.

Prerequisites

  • A Tenderly account (Free, Developer, or Pro plan)
  • At least one Tenderly project

Connect to the MCP Server

Run the following command in your terminal:

claude mcp add tenderly --transport http https://mcp.tenderly.co/mcp

Verify the server is connected:

claude mcp list

Run /mcp inside Claude Code, select tenderly, and follow the browser login to authenticate.

First tool call

Once connected, try a simple prompt to verify everything works:

“What is my account name and currently active project?”

Claude will call get_user_info and returns your account information and the currently active project (if set). If this works, you’re all set.

Set your active project

Most tools require a project context. Set it once per session:

“Set my active project to my-project under my-account”

Claude will call set_active_project. All subsequent tool calls (simulations, VNets, traces) will use this project.

You can also let Claude figure out the project automatically — just ask it to do something and it will call list_projects first if needed.

Troubleshooting

Authentication fails or the connector won’t log in

If the OAuth login loops or errors out, disconnect the Tenderly connector in your Claude client (Settings → Connectors) and re-add it. In Claude Code, run /mcp, select tenderly, and redo the browser login. Custom connectors on Claude.ai and Claude Desktop require a Pro or higher plan — on Enterprise plans, only workspace admins can add them.

”No active project” errors

Most tools require a project context. Ask Claude to set one:

“Set my active project to my-project under my-account”

Or let Claude pick automatically by asking it to list projects first.

”No active Virtual TestNet” errors

VNet tools operate on the active VNet. Either create/fork a VNet (which auto-activates it) or ask Claude to call set_active_vnet with an existing VNet ID.

Call trace or events are truncated

get_simulation_call_trace is capped at 256 entries and get_simulation_events at 100. For large transactions, use the Advanced Trace Navigation tools: start with get_trace_stats for size, then get_trace_skeleton to see the tree structure, and get_call_trace_node, get_subtree_events, or search_call_trace to drill into specific parts. For reverts, find_failures and get_error_path are usually the fastest path to the root cause.

Claude picks the wrong network, project, or VNet

Be explicit in the prompt — name the network, project slug, or VNet. You can also ask Claude to confirm the active context with get_user_info before running a task.

A tool call returns stale or wrong data

Active project and VNet state is per-session. If you switch projects or VNets mid-conversation, confirm the active context with get_user_info before running follow-up tools.

Next steps

OverviewTools Reference