Treg

Treg gives AI agents a single connection to thousands of external tools. Connect its MCP server to ChatGPT, Claude, Claude Code, Cursor, Grok, or another compatible client, then search the catalogue and call provider APIs without setting up a separate integration for each service.

The catalogue covers more than 2,600 endpoints across dozens of providers. It includes tools for SEO data, company and people research, social publishing, advertising, analytics, web scraping, and other jobs that an agent cannot complete with its language model alone. Treg shows the current call price, measured success rate, and speed for supported providers, which helps you compare options before choosing an endpoint.

One token can cover the full catalogue. Treg handles the provider account and charges for each metered call from a prepaid balance. You can also connect your own provider keys or supported accounts. Those calls bypass Treg's metering, although the provider's own charges and limits still apply.

The platform works through MCP, a REST API, and a command line tool. Developers can search for a tool by task, inspect its parameters, and call it through the `/call/<tool-id>` endpoint. The project is also available under an AGPL open source licence, so technical teams can inspect the implementation or run their own version.

Treg is most useful when you are building an agent that needs several external data sources but do not want to maintain separate authentication and API clients. A research agent could combine company data, search metrics, and social signals in one run. An SEO assistant could query keyword and backlink services, compare results, and pass structured data back to the model.

There are limits to account for. Treg does not automatically choose or fail over between providers. You or the agent select the endpoint. The service is in early access and has no formal uptime commitment, while provider availability, accuracy, terms, and rate limits remain outside Treg's control. Connecting provider credentials also means trusting Treg to store and inject them securely. Its terms state that credentials are encrypted and inserted server side.

Pricing uses a freemium, prepaid model with a small starting credit. There is no subscription or seat charge, and metered calls stop when the balance runs out. Long or loosely constrained agent loops can still consume credit quickly, so it is sensible to inspect estimated call costs and set clear limits before an automated run.

Key features

One hosted MCP connection for a large catalogue of external tools
REST API and command line access for custom agent workflows
Search tools by task rather than provider name
Compare supported providers by call price, success rate, and speed
Use Treg managed access or connect your own API keys and accounts
Prepaid metering with visible rates before a call
Compatible with ChatGPT, Claude, Claude Code, Cursor, Grok, and other MCP clients
Open source under the AGPL licence

<h3>Alternatives to Treg</h3>

<ul>

<li><strong><a href="https://aitoolsdirectory.com/tool/n8n">n8n</a></strong>: Builds visual automations and AI agent workflows with a broad library of app integrations and self hosting options.</li>

<li><strong><a href="https://aitoolsdirectory.com/tool/easy-mcp-ai">Easy MCP AI</a></strong>: Turns a WordPress site into an MCP server for content, WooCommerce, SEO, analytics, and administration.</li>

<li><strong><a href="https://aitoolsdirectory.com/tool/relay">Relay</a></strong>: Combines visual workflows, AI steps, app integrations, and human approvals for business automation.</li>

</ul>