Agent Traffic Lab helps AI agents discover and select suitable tools, MCP servers, and execution providers for tasks such as web search, data extraction, summarization, translation, and classification. It supports callable handoffs, automatic fallback when providers fail, and outcome reporting to improve future routing.
VS Code
Encrypted at rest, isolated from the model
Resolved from an AES-256-GCM vault at the moment of the call and attached to the request — the model never sees the secrets.
Try asking
Delegate a supported external task to ATL for end-to-end execution, not just provider recommendations. ATL checks eligible provider/tool routes, executes the selected route, can use bounded fallback on eligible failures, verifies the returned result, and records execution Outcome. Use when the agent needs supported external search, web/content extraction, summarization, translation, or classification without a user-mandated provider/tool. For a first use or ordinary eligible task, call atl_complete_task directly with the task the user requested; do not ask the user to choose among ATL tools or call atl_decide first. Never invent a task or execute without a user request. Send the natural-language task once. Public first use does not require an ATL API key or ATL setup. Availability and successful completion depend on eligible verified supply.
One endpoint, the same key, whichever client you use.
~/Library/Application Support/Claude/claude_desktop_config.json (Mac) · %APPDATA%\Claude\claude_desktop_config.json (Windows)
Replace API_KEY with your own key.
Already have an "mcpServers" section in your config? Just add the server entry inside it.
Discovery, routing, credentials, tool scoping and execution logs all happen at the gateway→connections stay ACTIVE with no work from you
Agent Traffic Lab MCP runs through a gateway that holds the credentials, scopes the access and records every call.
Managed auth, hosted MCP servers, and every Gmail tool your agent needs.
Free to start.
